<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[360Learning Engineering - Medium]]></title>
        <description><![CDATA[In this blog, the 360Learning Engineering team shares lessons learned from building the future of Collaborative Learning. - Medium]]></description>
        <link>https://medium.com/360learning-engineering?source=rss----79e2939dd800---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>360Learning Engineering - Medium</title>
            <link>https://medium.com/360learning-engineering?source=rss----79e2939dd800---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 09:48:49 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/360learning-engineering" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[leHack 2025]]></title>
            <link>https://medium.com/360learning-engineering/lehack-2025-0b765dbe43e2?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/0b765dbe43e2</guid>
            <dc:creator><![CDATA[Paul Van-Elsue]]></dc:creator>
            <pubDate>Fri, 25 Jul 2025 07:05:25 GMT</pubDate>
            <atom:updated>2025-07-25T07:05:25.170Z</atom:updated>
            <content:encoded><![CDATA[<p>Ever wondered what goes down at 3AM inside Paris’s Cité des Sciences… filled with hackers ?</p><p>From CTFs and bug bounty wins to payload crafting, beer-fueled workshops, and lockpicking challenges: step into the wild world of <strong>LeHack 2025 </strong>!</p><p>Check out this exclusive return of experience from my trip to <strong>France’s biggest technical hacking conference</strong> and what I brought back (besides stickers and hoodies).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Co9iSGl0LOsXsca9HyeXdA.jpeg" /><figcaption>HackTheBox meetup<br>Photo credit: <a href="https://x.com/OsintFr/status/1938530086036414792/photo/2">https://x.com/OsintFr/status/1938530086036414792/photo/2</a></figcaption></figure><h3>🏴‍☠️ LeHack 2025</h3><p>From June 26 to 28, I dove headfirst into <em>LeHack</em>, France’s flagship hacking event, right in the heart of Paris. It all kicked off with a HackTheBox meetup and carried on through two chaotic, caffeine-fueled days (and nights) of conferences, CTFs, bug bounty, and great conversations with the hacking community.</p><p>We were 8 hackers sharing one Airbnb: sounds like the start of a sitcom, but it was more like a LAN party meets <a href="https://defcon.org/">DEFCON</a>.</p><h3>🎤 Talks</h3><p>The schedule ran non-stop from 9 AM to 3 AM. I explored a variety of talks, including:</p><ul><li><strong>Active Directory pentesting</strong> (a major theme every year)</li><li><strong>Scammers</strong> and how they operate</li><li><strong>Minitel</strong> hacking</li><li>Demos of tools like <em>Nmap Viewer</em> and more</li></ul><p>The <a href="https://lehack.org/fr/2025/tracks/conferences">conferences list</a> is available here but this year, I focused more on hacking than attending talks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-N245xRaW8zsNGoa1vxgMg.jpeg" /><figcaption>The scene used for the talks<br>Photo credit: <a href="https://x.com/OsintFr/status/1938855117446619218/photo/1">https://x.com/OsintFr/status/1938855117446619218/photo/1</a></figcaption></figure><h3>🛠️ The stands</h3><p>The expo area was buzzing with <strong>public institutions, companies, schools</strong>, and of course, <strong>community groups</strong>.</p><p>Some highlights:</p><ul><li><strong>OSINT Industries</strong> gave out free trials of their intel platform and ran OSINT challenges where you could win a robot (I watched, didn’t play).</li><li><strong>YesWeHack</strong>, both a company and a bug bounty community hub, was everywhere: with their CTF, bug bounty, and of course, swag. I’ll talked more about them later one because I participated to both the CTF and the live BugBounty.</li><li><strong>OSINT-FR</strong>, a community booth, gathered the open-source intelligence crowd for knowledge-sharing.</li><li>A <strong>car hacking</strong> demo drew curious crowds.</li><li>There was even a “Who Wants to Be a Bitcoinnaire ?” game (yes, really) inspired by <em>Who Wants to Be a Millionaire</em>.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jfUZ7dHAljW2-XcR22HhAA.jpeg" /><figcaption>Car hacking<br>Photo credit: <a href="https://lehack.org/files/062422-121456-e1750325502154-1024x1024.jpg">https://lehack.org/files/062422-121456-e1750325502154-1024x1024.jpg</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*axDcDwKX7Rpzqu4HP4Vfcg.jpeg" /><figcaption>Lockpicking<br>Photo credit: <a href="https://x.com/OsintFr/status/1938616139598233922/photo/2">https://x.com/OsintFr/status/1938616139598233922/photo/2</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hPpsXu-O3gWjzzaiiwq08w.jpeg" /><figcaption>OSINT Industries’s robot<br>Photo credit: <a href="https://x.com/OsintFr/status/1938616139598233922/photo/1">https://x.com/OsintFr/status/1938616139598233922/photo/1</a></figcaption></figure><h3>🧠 Workshops</h3><p>Workshops were also part of the experience. I joined one focused on <strong>industrial equipment hacking</strong>: fascinating, but I’ll admit it was a bit above my level. Still, a great glimpse into a niche yet critical part of cybersecurity.</p><h3>🔥 One Payload to Rule Them All: YesWeHack CTF</h3><p>The <strong>YesWeHack CTF</strong> was refreshingly original and so I participated actively to this one. The goal ? Exploit <strong>13 different web vulnerabilities</strong> using a <strong>single payload</strong>.<br>For those less technical: a <em>payload</em> is what you inject to trigger a vulnerability (think: SQL injection, XSS, etc.).</p><p>After two long nights (read: mornings), I landed <strong>10th out of 255</strong>. Ahead of some heavy-hitters in the bug bounty scene.<br>Here’s my final payload:</p><pre>&#39;/*{$smarty.server.FLAG}{*[.&gt;]&lt;[,.]&#39;UNION SELECT flag,1,1 FROM flag--*/UNION SELECT flag,1,1 FROM flag--;&lt;/script&gt;&lt;script&gt;alert(flag)&lt;/script&gt;[&quot; or 1=1]/password/text(){% if lipsum is not defined %}{{[&#39;FLAG&#39;]|map(&#39;getenv&#39;)|join}}&lt;%= ENV[&#39;FLAG&#39;] %&gt;{% endif %}{literal}{{lipsum.__globals__.os.environ.FLAG}}{/literal} *}&#39;; echo $FLAG+&#39;../../../../../../../../../../proc/self/environ</pre><p><strong>For you: It’s long. It’s ugly. For hackers: it’s art, it’s beautiful !</strong> Mixt of bash, SQL, javascript and others, every single char is important to escape the other langages, don’t create a bug on another vulnerability and still exploit the one you’re working on.</p><p>By being on the leaderboard, I was rewared with a hoodie, a beanie, and a t-shirt.. and a big smile.</p><h3>💸 My First Bounty: Broken Access Control</h3><p>Saturday was live bug bounty day. My team of 5 hunted vulnerabilities on a real-world app.<br>We found a <strong>broken access control</strong> issue that allowed access to customer RIBs by modifying a cookie via a separate route. The result?<br>👉 One valid bug.<br>👉 <strong>€400 reward (to be shared within the team)</strong>.<br>👉 My <strong>first-ever bounty</strong>.</p><p>I’m really glad to get my first bounty, not because of the money but because it’s a hard exercise and for the first time I succedded. I hope it’s the first one of a longue serie !</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_DwZia72FO30S2DVOxHZFw.jpeg" /><figcaption>Me at the Live Bug Bounty (the only one who looks at the camera)<br>Photo credit: a friend of mine</figcaption></figure><h3>🤝 Community: The Real Power-Up</h3><p>Beyond the tech, tools, and talks, what really stood out was the <strong>spirit of collaboration</strong>.<br>Cybersecurity evolves fast. It thrives on <strong>shared knowledge</strong>, curiosity, and learning from others.<br>That’s what LeHack is all about: helping each other grow, challenge, and inspire.</p><h3>Cybersecurity isn’t just for experts: it’s for the curious.</h3><p>Events like LeHack remind us how exciting, creative, and collaborative this field can be. Whether you’re into solving puzzles, building things, breaking them (ethically !), or just learning how it all works, there’s something in cyber for everyone.</p><p>So if any of this sparked your interest, come chat. We’re always happy to share, learn, and explore together !</p><p>On my side, I’m already waiting for the next edition !</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0b765dbe43e2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/lehack-2025-0b765dbe43e2">leHack 2025</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How we are making Accessibility a reality at 360Learning]]></title>
            <link>https://medium.com/360learning-engineering/how-we-are-making-accessibility-a-reality-at-360learning-80ce0edad5b9?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/80ce0edad5b9</guid>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[accessibility]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Thibault Brazillier]]></dc:creator>
            <pubDate>Fri, 25 Jul 2025 07:05:15 GMT</pubDate>
            <atom:updated>2025-07-25T07:05:15.271Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*fEWgcvxXl_38Si9i" /><figcaption>Photo by <a href="https://unsplash.com/@sigmund?utm_source=medium&amp;utm_medium=referral">Sigmund</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Getting started</h3><h4>Context</h4><p>By mid-2023, it became clear that accessibility was a growing challenge for our platform. As our company began targeting larger organizations, accessibility moved from a secondary concern to a critical issue, sometimes even a legal requirement. We faced increasing pressure from clients and prospects who demanded a higher standard. Moreover, we recognized internally the need to make our product accessible to all, including future employees with disabilities.</p><p>However, our platform was far from meeting industry standards, and accessibility had not been a focus in our development process. The challenge was not just about fixing immediate issues; it was about fundamentally rethinking how we design and build our product.</p><p>Our team, the Developer Experience Frontend team, was tasked with addressing this complex project. Our squad provides support for the developers of around 10 teams on frontend topics and tooling. And we also led transversal projects on the frontend codebase which was the case for this Accessibility initiative.</p><h4>Task force and goals</h4><p>Accessibility is a large-scale project that involves the whole EPD (Engineering, Product, Design) team as features must be created with accessibility in mind throughout the whole process. It is also a new standard we want to enforce for the long term. It’s not just about fixing current issues but about spreading knowledge and raising awareness, making accessibility second nature for everyone going forward.</p><p>To tackle this ambitious and transversal goal, we formed a task force with members from all three parts of the EPD team. We began by establishing clear goals, based on our research of the industry standards and feedback from our existing customers. Our current goal is to improve the accessibility of the product in order to later conduct an accessibility audit to achieve <a href="https://www.w3.org/TR/WCAG21/">WCAG 2.1</a> compliance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zutJQ5L5Wl97n9RlETuhHw.png" /><figcaption><a href="https://www.w3.org/WAI/images/WCAG21/small-1024x512@1x.png">https://www.w3.org/WAI</a></figcaption></figure><h3>3 Levers</h3><p>To achieve this, we focused our action on 3 levers:</p><ul><li>building an accessibility culture in the EPD team</li><li>writing guidelines and processes to enforce accessibility</li><li>improving our product and codebase.</li></ul><h4>Evangelizing in the team</h4><p>Building an accessibility culture begins with making the team aware of the challenges users face when interacting with our product. We had the opportunity to do so during one of our All-Hands meetings where we presented the accessibility project. We presented general figures and statistics but also showed the team how poorly our product fared in terms of accessibility. Our presentation revealed that our basic use cases were inaccessible with just a keyboard and screen reader, which came as a wake-up call for the company. The feedback was positive, motivating us to push forward.</p><p>Another great event to get people talking and thinking about accessibility was the <a href="https://accessibility.day/">Global Accessibility Awareness Day</a> (GAAD) which occurred on May 16th. We decided to organize a workshop day, with team-specific workshops in the morning for the design, product, and engineering teams and an inclusivity workshop in the afternoon. In these sessions, we tackled accessibility issues practically, with task force members leading as subject-matter experts. Through this initiative, we aimed to further raise awareness about the importance of accessibility while also sharing knowledge and practical skills to empower 360Learners in their specific roles.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xbXufHYtFouWvXCnGM0Fzg.png" /><figcaption><a href="https://accessibility.day/">https://accessibility.day/</a></figcaption></figure><h4>Crafting guidelines</h4><p>Making the team aware of accessibility issues wasn’t quite enough, we also needed to give them the knowledge of how to tackle those issues. After researching the topic in depth using resources such as the <a href="https://www.w3.org/WAI/">World Accessibility Initiative</a> or the <a href="https://developer.mozilla.org/en-US/docs/Web/Accessibility">Modzilla Developer Network</a>, we developed courses on our platform, starting with a general course on accessibility and its impact on users. We also created detailed courses to help developers understand common accessibility issues on our platform and key web accessibility concepts.</p><p>Alongside our efforts on the development side, we created guidelines for designers. These guidelines help them use our generic components better and ensure the User Experience remains accessible. With those, we wanted to build a solid foundation in both functions to ensure that all new features would be designed and implemented in an accessible manner from start to finish.</p><h4>Improving the product</h4><p>The last pillar of our action was to bring concrete improvements to our product and its existing features. To help us with this task, we partnered with <a href="https://www.levelaccess.com/">Level Access</a>, an expert in digital accessibility. They help us identify current accessibility issues on our platform and offer advice on design drafts for new features. They also audit our platform to deliver a <a href="https://en.wikipedia.org/wiki/Voluntary_Product_Accessibility_Template">VPAT</a>, the standard document to assess the web accessibility of a product.</p><p>We started by defining some key user stories in our product to focus our efforts on. While making the entire product accessible is our ultimate goal, we had to prioritize due to limited resources. We decided to focus on high-impact changes by targeting the most critical user stories.</p><p>Once identified, Level Access audited those scenarios, giving us a clear understanding of the accessibility problems across those user journeys and how to address them. To fix them, we sorted them and dispatched them among our product squads which enabled us to lighten our load as well as spread the knowledge so that squads would slowly become autonomous in their handling of accessibility issues. Level Access continues to support us on this journey. They provide advice on our proposed fixes and validate them with their in-house testers and experts.</p><p>In order to fix accessibility at scale, we relied on our framework of generic components. Those components are used all over the product to bring design consistency. By making those components accessible with resources like the <a href="https://www.w3.org/WAI/ARIA/apg/patterns/">WAI patterns</a>, we were able to bring a base level of accessibility all over the platform. When you have more context about the concrete use of the component, it’s important to provide more information to the user using technologies such as ARIA. However relying on those components was a high-impact way for us to address accessibility issues throughout the product.</p><h3>What’s next</h3><h4>Current state of the project</h4><p>As of now, we are in the process of fixing the identified accessibility issues before the VPAT audit of Level Access. This has been a lengthy project, but we’ve seen significant improvements in both the platform and the team’s mindset and awareness around accessibility. It’s hard to measure how close we are to a fully accessible product. However, given our starting point, we’ve made significant progress, and it’s already yielding benefits like smoother keyboard navigation, which improves usability even for regular users.</p><h4>Perspectives</h4><p>Building accessibility in a product is not a one-time endeavor but a commitment to principles and practices that need to be reviewed and audited continuously in the future. Our current target is the VPAT, and we’ve concentrated our efforts on critical user stories. We believe our efforts have laid a strong foundation, setting us on the path to becoming an accessibility-first company, committed to continuously improving our product’s accessibility.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=80ce0edad5b9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/how-we-are-making-accessibility-a-reality-at-360learning-80ce0edad5b9">How we are making Accessibility a reality at 360Learning</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Unicorns&Cookies 2]]></title>
            <link>https://medium.com/360learning-engineering/unicorns-cookies-2-ac215c4348f8?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/ac215c4348f8</guid>
            <category><![CDATA[startup-life]]></category>
            <category><![CDATA[sharing]]></category>
            <category><![CDATA[meetup]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Guillaume Vibert]]></dc:creator>
            <pubDate>Mon, 07 Jul 2025 15:29:25 GMT</pubDate>
            <atom:updated>2025-07-07T15:29:25.367Z</atom:updated>
            <content:encoded><![CDATA[<p>We Bring the Conference Directly To You</p><p>When I sent the invitations for this year’s U&amp;C, I got some questioning looks from the newcomers who were not there <a href="https://medium.com/360learning-engineering/unicorns-cookies-42b34a656d0b">last year</a>. I must admit that the name of the event is not self-explanatory 😅- so let me reintroduce it:</p><p>We work in a wonderful community where people are eager to share what they work on and help everyone grow. At 360Learning, where Upskilling is our motto, we invited speakers from across the French Tech to speak to us on a large variety of topics — with the additional benefit of sharing drinks and cookies. Here is a quick debrief on the four talks.</p><figure><img alt="A cute unicorn lifting cookies and the 360Learning logo" src="https://cdn-images-1.medium.com/max/1024/1*GbzhXt4esepEqDdbWn3PlQ.png" /><figcaption><em>Our new mascot! Instead of AI-generating one like last year, we did an internal drawing contest and FullStackie was born!</em></figcaption></figure><p><a href="https://www.linkedin.com/in/theotimeleveque/">Théotime</a> from Back Market: <strong>Staff Engineer, The Challenges, The Headaches, and How to Overcome Them</strong></p><p>How lucky were we to have a talk straight from the <a href="https://www.youtube.com/watch?v=EiADjypQoY0&amp;t=1264s">DevOxx</a>. The career of a software engineer forks at some point in their seniority. Some get to managing, some stay in a pure individual contributor role, and some evolve to a staff or principal engineering role. Contrary to the move from Junior to Senior, it’s not a straightforward progression; new skills to be unlocked, and new expectations to answer to. To help us achieve such a move, Théotime brought some very valuable insights, like making oneself identifiable, triggering listening situations, and not spreading too thin… I strongly recommend watching the talk and coming back to it every now and then. It takes time to digest and incorporate into a daily habit. Plus, join Staff42 to exchange with other staff engineers!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TVHr30_YFSmF_bSyUOilLA.jpeg" /></figure><p><a href="https://www.linkedin.com/in/raphaeltahar/">Raphael</a> from Ledger: <strong>The V8 Engine</strong></p><p>I found out Raphaël’s impressive résumé while browsing Staff42’s list of members. From there, I spent some time reading his <a href="https://medium.com/@tahar.raphael">Medium</a> publications and thought that if he was able to get them from paper to speak, that would be the Pure Tech Talk of the night. And did he deliver! A deep dive into the JavaScript V8 engine that runs the internet, the difference between interpreting and compiling, some direct actionables in both front and back-end code to improve performance — and a great reminder that improving performance has a direct impact on the health of our planet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DQOARpBR_QU1kI7f2GvARg.jpeg" /></figure><p><a href="https://www.linkedin.com/in/samirabechar/">Samira</a> &amp; <a href="https://www.linkedin.com/in/christelle-leroy/">Christelle</a>: <strong>DEI and its impact in Tech</strong>.</p><p>Did you know that facial recognition software (yes, the kind of tech that enables your phone to recognize you) fails 35% of the time on Black Women, 1% on White men? That 40% of people are victims or witnesses of discriminatory acts in the enterprise every year? That 50% of women in tech leave the industry before 35 years old? Samira and Christelle brought some of those chilling facts to our attention and gave us concrete <a href="https://www.amazon.fr/Yeeso-Lavenir-techForAll-Acc%C3%A9l%C3%A9rer-f%C3%A9minisation/dp/B0F92XG5X3">actions</a> we could do to change the world, a little bit.</p><p><a href="https://www.linkedin.com/in/eliotandres/">Eliot</a>: <strong>Training an image generation model and deploying it to millions of users</strong></p><p>360Learning moved early this year, and our new offices are now located very next door to another fellow FrenchTech superstar, Photoroom. CTO Eliot agreed to do the trip, and talked to us about training and deploying an AI model. An insightful AI talk, very approachable but technically deep enough to excite our MLEs. It’s a pleasure to see that the French Tech can produce what could be reserved for the depths of Silicon Valley.</p><figure><img alt="Yummy cookies. I ate too many of those but I don’t care" src="https://cdn-images-1.medium.com/max/1024/1*5aYMfKFDo_iWax3T-G981w.jpeg" /><figcaption>Some cookies</figcaption></figure><p>To be honest, organizing such an event is an exhausting commitment, but the people met, the connections created, the quality of the talks delivered, and the enthusiasm of the audience make it worthwhile. Thanks again to all the speakers who came — I guess I will have to explain again next year what a Unicorn &amp; Cookie is…</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac215c4348f8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/unicorns-cookies-2-ac215c4348f8">Unicorns&amp;Cookies 2</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Unicorns&Cookies]]></title>
            <link>https://medium.com/360learning-engineering/unicorns-cookies-42b34a656d0b?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/42b34a656d0b</guid>
            <category><![CDATA[sharing]]></category>
            <category><![CDATA[meetup]]></category>
            <category><![CDATA[conference]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Guillaume Vibert]]></dc:creator>
            <pubDate>Tue, 04 Jun 2024 06:58:48 GMT</pubDate>
            <atom:updated>2024-06-05T12:50:04.181Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Unicorns in suits sit on cookies — AI-generated image" src="https://cdn-images-1.medium.com/max/1024/1*MxlrbvLy6zmQDlLxx4aU8Q.jpeg" /></figure><p>At 360Learning, as a truly full remote company, we organize events to get together and build up team spirit. One of those is the monthly <a href="https://medium.com/360learning-engineering/sharing-fun-times-and-knowledge-beer-n-tech-at-360learning-fa4f1bf75ff9">Beer n’ Tech</a> for Engineering, Product &amp; Design. To begin the year, we wanted to make something special, bigger, better.</p><p>At the same time, we just went through a major shake-up in the organization of the Engineering team. From one global Engineering Manager with 80 developers, we have moved to 15 squads, each with a dedicated Tech Lead. This is a new role, for the company, but also for most of the new Leads. This pivotal job is very different from a Senior Software engineer position that most had before and with a high potential of messing up both oneself and the whole team.</p><p>Those two situations gave us the idea of organizing a very special Beer n’ Tech, one where we could ask our peers from other, maybe more mature companies, to come to share with us what the job of Tech Lead could bring, what organization challenges they face and in general, how to make an impact as a tech in the company.</p><p>So we reached out to our network — and asked for quotations from different bakeries.</p><h3>Julien</h3><p><a href="https://www.linkedin.com/in/julientanay/">https://www.linkedin.com/in/julientanay/</a></p><p>Julien is an Engineering Manager in the “Engineering Enablement” department of Doctolib. Doctolib distributes an online appointment booking service for patients in Europe. Julien supports diverse teams of SWE and SRE in charge of everything related to developer happiness and productivity, from their local dev env to their CI platform, from reusable code bits to shared and/or automated workflows.</p><p>Julien invited us to rewind together the last couple of years, across different companies and cultures, when he traded technical leadership for people management. He learned more about the job — and himself — in 3 years than in his entire career and, plot twist! He’s going back to an Individual Contributor track this year.</p><p>First, he introduced us to how Doctolib is organized with teams of 4–6 Fullstack devs, 1 Engineering Manager, 1 Product Manager, 0.5 UX Designer, and 0.5 Product Data Analyst. He highlighted the two different tracks between full-time management and ICs, where no hybrid role exists.</p><p>After highlighting the huge difference between technical leadership and engineering management, he shared different solutions for the organization to prepare people for their transition from tech expert to full-time managers:</p><ul><li>Raise awareness about the EM role and create a plan for aspiring EMs</li><li>Create and maintain your career framework, and define boundaries</li><li>Support the “EM team” through regular trainings and reflection groups</li><li>Provide HR support</li></ul><p>But most importantly:</p><ul><li>Tell everyone it’s ok to go back from full-time manager to tech expert</li></ul><p>Management is a very different job and people can discover it’s not made for them. They shouldn’t feel stuck in it. It will also reassure everyone who aspires to become EM.</p><p>It was refreshing to attend this rare presentation about moving back from management. Thank you Julien for genuinely sharing your own experience!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mjd1SbvGg1sNE28EOtQdsw.png" /></figure><h3>Simon</h3><p><a href="https://www.linkedin.com/in/simondolle/">https://www.linkedin.com/in/simondolle/</a></p><p>Simon is an Engineering Manager at Contentsquare, a French unicorn that helps businesses optimize and improve their users’ digital experience.</p><p>Simon shared with us the challenges of growing engineering teams and how they’ve been iterating on their organization to face these challenges. Some notes from Simon’s talk:</p><p>Engineering teams at Contentsquare are composed of tech leads and a manager and have rituals such as dailies, sprints, and retrospective meetings.</p><p>But as the teams started to grow, certain rituals required modification, like the division of daily meetings, and the introduction of new ones, including tech leads meetings and cross-department/R&amp;D meetings. They also established a “CTO Office” to ensure Transversal synchronization.</p><p>These changes didn’t entirely resolve certain issues such as the management of inter-team dependencies. To combat this “dependency hell”, Simon advocates for larger and mixed teams.</p><p>His final takeaways:</p><ul><li>Organization depends on company scale and business</li><li>Skills barriers add a lot of friction</li><li>Alignment requires a dedicated organization</li></ul><p>Thank you Simon for openly sharing the struggles you’ve been through — there were great lessons for us to learn from. And we were absolutely captivated by the illustrations in your presentation slides!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/959/0*qxtxe8BhdFE7lUrw" /></figure><h3>Hugo</h3><p><a href="https://www.linkedin.com/in/hugolassiege/">https://www.linkedin.com/in/hugolassiege/</a></p><p>Our most <a href="https://eventuallycoding.com/">famous</a> speaker had a global presentation on how a software engineer should make an impact on their company. Too often important discussions can be done between top managers — without the tech in the room. (Personal note, I have definitely observed that, especially in companies where technology is not in the DNA, where it may have been introduced a couple of decades ago as a follow-up to industrialization). Getting into leadership is a way to make sure that we are in the right place.</p><p>There were a lot of insights to unpack. I won’t do a catalog, you can check Hugo’s <a href="https://www.youtube.com/@eventuallycoding">resources</a>, here are some of my favorites:</p><ul><li>Create friction, but be the problem-solving person</li><li>Inclusion: it should not be an us vs them mentality.</li><li>Climb up the ladder by having an impact on a bigger range: your squad, several, the whole team, the company, the industry!</li></ul><p>I especially liked Hugo’s vision of having tech people as CEOs, not “just” CTOs. Tech companies would profit from having tech people making the biggest decisions.</p><h3>The event</h3><p>In the end, we had:</p><ul><li>Great speakers,</li><li>Great attendance: Despite the holidays, we had ¾ of the engineering team show up, so the subject did not interest only the Tech Leads,</li><li>Great cookies and apéro,</li><li>And a great time! We’ll do something like that again next year, you still have time to <a href="https://jobs.lever.co/360learning">join</a> us for the second edition!</li></ul><figure><img alt="cookies!" src="https://cdn-images-1.medium.com/max/1024/1*-dczArN_u6I0BGl24bSfHw.png" /></figure><p>Written by <a href="https://medium.com/@fabrice.gaudin">Fabrice Gaudin</a>, <a href="https://medium.com/@guillaume.vibert">Guillaume Vibert</a> and <a href="https://medium.com/@marie.douriez">Marie Douriez</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42b34a656d0b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/unicorns-cookies-42b34a656d0b">Unicorns&amp;Cookies</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Sharing fun times and Knowledge: Beer n’ Tech at 360Learning]]></title>
            <link>https://medium.com/360learning-engineering/sharing-fun-times-and-knowledge-beer-n-tech-at-360learning-fa4f1bf75ff9?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/fa4f1bf75ff9</guid>
            <category><![CDATA[sharing]]></category>
            <category><![CDATA[meetup]]></category>
            <category><![CDATA[conference]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Guillaume Vibert]]></dc:creator>
            <pubDate>Wed, 10 Jan 2024 09:25:06 GMT</pubDate>
            <atom:updated>2024-01-10T09:25:05.994Z</atom:updated>
            <content:encoded><![CDATA[<p>A little bit of honesty to start: I have not invented this principle, quite the opposite. In one of my previous experiences, this was a usual thing: people would book one hour to present a technical subject of their choosing, with beers, chips, and sodas to share. In an R&amp;D of 600 persons, this was a monthly, or bimonthly opportunity to present the result of a project, or sometimes a good book or conference. I fell in love with the concept with one specific occurrence, where a Data Scientist came and went completely unexpected, by presenting one of her previous lives, where she was an astronomer. That was a breath of fresh air, learning about black holes and the life of those scientists, working around the world, looking for stars and credits.</p><p>When I joined 360Learning, I was delighted to see that this was already kind of a thing: every other week, we had an engineering all-hands spot where one squad would share with the whole team whatever they wanted. Sadly, I discovered that the focus was very much restricted to the status of the OKR, and the hardcoded rhythm meant that you had to present when it was your turn and wait almost a year when it was not. There was room for a Beer n’ Tech concept, but would people be interested, or would I sit alone in a room with M&amp;Ms and a mic?</p><p>One year after the first edition, I’m glad to be able to say that it worked. We now have a monthly team session with 2 different sets of speakers and a packed room every time. How did we get there, and why should you as well invest in your own version of the concept?</p><h3>Finding Speakers</h3><p>The most important part of the Beer n’ Tech. People are busy, people are shy, people (think they) don’t have anything interesting to talk about. I rarely have spontaneous applications to present. When I do, it’s from teammates with experiences talking at conferences or meet-ups (think data scientists and ML engineers, team leaders, or C-level). Therefore, I hunt for speakers.</p><p>I try to make it clear what the expectations are: We are not dealing with career-threatening presentations. Speakers should spend only a few hours on their presentations with light slides (the audience came to listen, not to read). We propose a rehearsal session a few days before the date, with immediate feedback and a chance to improve the slides and evacuate the stress.</p><h3>Having a good time</h3><p>360Learning is a <a href="https://360learning.com/blog/360learner-journey-remote/">remote-first company</a>, but with a nice office in Paris where 20 to 30 360Learners can easily meet up. Bringing food is not a high budget (think 60€ per event) and it works. The talks end at 6 PM, but the party goes on while there is still something to drink, with a chance of further prolongation at the closest bar. It’s less fun for the folks who hear us remotely, but that’s a blessing in disguise: we often see teams using their monthly team-building budget to spend the day together and cap it by the BnT session.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Was9JJDwtjCWH0qc" /></figure><h3>Making it Special</h3><p>All of our sessions have a special place in my heart, but there are some occurrences where we did something different :</p><ul><li>After joining our company, <a href="https://360learning.com/blog/new-advisor-guillaume-vives/">G</a> traveled from California to spend a few days talking with executives. We took this opportunity to have him at a Beer n’ Tech present himself and his career in a very chill way. Sharing drinks and fun anecdotes was a great way to break any intimidation his impressive career could have given at first sight.</li><li>When we acquired <a href="https://360learning.com/blog/360learning-elamp-acquisition/">eLamp</a>, we first welcomed our dozen new teammates with a bunch of friendly emotes on Slack — but the next BnT was the real opportunity to both know their journey and product and know them personally.</li><li>Through some random opportunities, I found out that two of my teammates had wine-producing families. Just before Christmas, they took the stage to mix presenting their teams and have us taste their production (ie the day we did a Wine n’ Tech).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LFqDmDeEpxcPEv43" /></figure><p>Here’s a sample of the diversity of other subjects that were addressed in this year’s Beer n’ Tech :</p><ul><li>Pure technical presentations (introducing KraphQL or Kotlin and a 2 parts on the benefits and dangers of a microservice architecture)</li><li>End of projects celebration (<a href="https://360learning.com/blog/product-update-ai-question-generator-enhancements-customization/">QGen</a> and <a href="https://www.youtube.com/watch?v=woqw7sazMAw">Dashboard</a>)</li><li>Slice of life (Maxime is a former client of 360Learning, now part of the team and Rapha gave us a hilarious tell of his life in the Middle East)</li></ul><p>While I hope hosting a Beer n’ Tech will never be mandatory, I’m glad to see that it is now rewarded. Management has added a soft skill in t<a href="https://360learning.com/blog/360learner-journey-engineer-career-path/">he career path</a> to recognize folks who go out of their way to do a Beer n’ Tech and share with their peers. After all, we are the Collaborative Learning company, Collaborative Teaching should be in our DNA.</p><p>Do you have a similar experience at work? Interested in launching such an initiative? Let’s share some tips 🍻!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa4f1bf75ff9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/sharing-fun-times-and-knowledge-beer-n-tech-at-360learning-fa4f1bf75ff9">Sharing fun times and Knowledge: Beer n’ Tech at 360Learning</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an AI-powered Question Generation feature at 360Learning, before and after ChatGPT]]></title>
            <link>https://medium.com/360learning-engineering/building-an-ai-powered-question-generation-feature-at-360learning-before-and-after-chatgpt-5a1823ec7feb?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/5a1823ec7feb</guid>
            <category><![CDATA[generative-ai-use-cases]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Marie Douriez]]></dc:creator>
            <pubDate>Mon, 17 Jul 2023 08:49:58 GMT</pubDate>
            <atom:updated>2023-07-17T08:49:58.818Z</atom:updated>
            <content:encoded><![CDATA[<p><em>In this blog post, we’ll explore how we implemented a </em><strong><em>Question Generation feature</em></strong><em> </em><strong><em>(QGen) </em></strong><em>for content creators on the 360Learning platform, </em><strong><em>before and after ChatGPT</em></strong><em>.</em></p><p>360Learning is a <strong>collaborative Learning Management System</strong> that <strong>streamlines the course creation process</strong> for internal L&amp;D and subject-matter experts. In June 2022, we started thinking about how we could leverage <strong>Artificial Intelligence (AI)</strong> to help our <strong>content creators</strong> create high-quality courses quickly and efficiently. After running several customer discoveries, we identified that <strong>devising</strong> <strong>relevant comprehension questions</strong> to test how well learners have understood the content of a course was a tedious task for content creators but essential to produce engaging content for learners.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RJGlykvFPGUxNS6Wkfn-sA@2x.png" /><figcaption>Tailored Content, Faster: Introducing QGen, Powered by Generative AI (<a href="https://360learning.com/blog/introducing-qgen-powered-by-generative-ai/">https://360learning.com/blog/introducing-qgen-powered-by-generative-ai/</a>)</figcaption></figure><blockquote>We identified that <strong>devising</strong> <strong>relevant comprehension questions</strong> to test how well learners have understood the content of a course was a time-consuming and tedious task for content creators but essential to produce engaging content for learners.</blockquote><h3>I) Building QGen with open-source models</h3><p>In the summer of 2022, ChatGPT was not a thing yet. To build our question generation feature, we leveraged some open-source models that could be found on <a href="https://huggingface.co/">HuggingFace</a> (a platform that hosts models and data).</p><h4>I.a) Generating questions</h4><p>To be most helpful to our authors, the question generation feature we envisioned had to generate <strong>pertinent questions</strong> regarding the course creator&#39;s input text and provide the <strong>correct answer </strong>for each question. To that end, we built a chain of two models:</p><ol><li>The first model is an “end-to-end” question generation model that generates question prompts from a text (without having to provide a specific answer). We used a t5-base model, already finetuned on that generation task using a public English reading comprehension dataset: <a href="https://huggingface.co/valhalla/t5-base-e2e-qg">valhalla/t5-base-e2e-qg</a>.</li><li>The second model is a question-answering model that takes as inputs the text and a previously generated question and extracts the answer to that question from the text. We used <a href="https://huggingface.co/deepset/roberta-base-squad2">roberta-base-squad2</a>.</li></ol><p>To offer some diversity in the types of questions that were suggested, we also generated <strong>Yes/No questions</strong> by following a similar logic: we chained an answer-aware question generation model (where the provided answer would be “Yes” or “No”) and a boolean question-answering model, to confirm whether the correct answer is Yes or No.</p><p>To reach a good <strong>quality</strong> and <strong>diversity</strong> of questions while maintaining a reasonable <strong>latency</strong>, there were a few modeling parameters to tweak, the most important ones being the <strong>number of beam</strong>s (see <a href="https://huggingface.co/blog/how-to-generate#beam-search">this article</a> from HuggingFace on beam search) and the <strong>length of the text chunks</strong> to process (a T5 model can theoretically ingest up to 512 tokens but the training contexts used for finetuning tended to be shorter).</p><p>Inspired by other optimization stories, we also found that using the <strong>ONXX</strong> format was able to speed up computations (<a href="https://huggingface.co/docs/optimum/v1.2.1/en/onnxruntime/modeling_ort">Optimum Inference with ONNX Runtime</a>). We also implemented a <strong>static cache</strong> with some of 360Learning’s frequently used cheat sheet templates.</p><p>On the processing side, some <strong>pre-processing</strong> steps were necessary to handle the input text that was in raw HTML, detect the input language (this first version of the model would only work for English) as well as prevent generation when sensitive words were detected (to that end, we used a list of “bad” words, also used for preprocessing the training data in the original T5 paper). Finally, we added <strong>post-processing</strong> steps that would discard redundant questions (caused by a higher number of returned questions in the generation phase) and questions deemed as less relevant (based on the score provided by the question-answering model in the generation phase).</p><h4>I.b) Testing the quality of the generated questions</h4><p>To assess the quality of the generated questions and whether they’d be relevant to our authors, we developed an <strong>internal testing process.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1dHNrQFv4ZdlnLAX_iXTrw.png" /><figcaption>Asking internal testers to rate some generated questions</figcaption></figure><ul><li>First, we sampled a few hundred sheets from our customers’ base and generated questions offline.</li><li>Second, we displayed these sheets and the corresponding generated questions using a Streamlit frontend, connected to a Google sheet to save results.</li><li>Third, we recruited and asked internal testers to review questions. Each question could be either “accepted” if the tester thought that, with some manual modification, the question could be reused in the course, or “rejected” if and only if the tester considered that this question was not at all reusable in your course.</li><li>Finally, we computed two metrics: % of rejected questions and % of sheets with 0 accepted questions. We compared these results to the minimum levels we had defined before launching the test for productization.</li></ul><p>Luckily, our models for English QGen outperformed our minimum criteria and received <strong>very positive internal feedback</strong>.</p><h4>I.c) Designing the production service</h4><p>To power our platform with QGen, we built an external service, using <strong>FastAPI</strong>, which we hosted on <strong>Kubernetes</strong> instances with CPUs.</p><p>After several iterations on the product workflow and to limit users’ waiting times, we conceived a feature where authors would not have to click on a button to trigger the generation of questions: questions would be generated in the <strong>background</strong> and would be automatically displayed upon opening the “Add a question” window. To manage such a system on the engineering side, we designed a <strong>queuing mechanism using Redis</strong> (and the Python library RQ).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Q8t7GW5ON-h0vzBS" /><figcaption>The QGen service architecture uses a queuing mechanism to support the background generation of questions.</figcaption></figure><h4>I.d) Expanding the feature</h4><p>This feature was released in production in <strong>February 2023</strong>, to companies who opted-in (see <a href="https://360learning.com/blog/introducing-qgen-powered-by-generative-ai/">our communication</a>).</p><p>Soon after, we extended QGen to <strong>videos</strong>, by leveraging automatic transcripts (this speech-to-text functionality was already implemented in the platform). On the modeling side, we explored ways to extend QGen to other languages, especially <strong>French</strong>. This turned out to be <strong>more challenging than expected</strong> for a couple of reasons: there were fewer open-source French models similar to the ones we had used in English and fewer good-quality datasets available that were under no commercial license. We did finetune some models internally but could not reach satisfactory results. That’s where <strong>ChatGPT</strong> came in very handy.</p><h3>II) Improving QGen with ChatGPT</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tLfZCP3a1Dx244XzNm21ww.png" /><figcaption><a href="https://openai.com/blog/chatgpt">https://openai.com/blog/chatgpt</a></figcaption></figure><p>You know the story: in December 2022, ChatGPT was released by OpenAI. It showed remarkable capabilities at generating all sorts of texts, ranging from poems to code and blog posts. We quickly grasped the opportunity to <strong>improve the quality of our generated questions</strong>: ChatGPT could indeed generate questions in <strong>multiple languages</strong>, provide new <strong>types of questions</strong> (questions with multiple answers for instance) as well as generate <strong>wrong answers</strong> for each question.</p><p>The first step to implementing a new version of QGen with ChatGPT was to ensure the protection and confidentiality of our customers&#39; data. Luckily, the Azure OpenAI Service provides <strong>secure access to OpenAI’s models</strong> (<a href="https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy">Data, privacy, and security for Azure OpenAI Service</a>).</p><p>After receiving approval from our security and legal teams, we started engineering a <strong>prompt</strong> that would take a text as input and generate questions in the language of the text, of multiple types (True/False, one-single answer, multiple choice) and with wrong answers. We followed the same testing process as with the first version and validated internally that the questions would be pertinent to our users. The results came out very positive!</p><p>Execution-wise, we were able to <strong>re-use most of the previously implemented bricks</strong>: the internal testing app, the product frontend workflow, the external service running on Kubernetes, and the backend communication between the service and the platform. With one major update: since the computations were now IO-bound and no longer CPU-bound, we <strong>updated our Redis queuing mechanism</strong> so that we could run hundreds of jobs in parallel (and replaced RQ with Arq).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3Ex61CfJC3R7cFXLTtd-KQ.jpeg" /><figcaption>QGen, powered by ChatGPT, generates three different question types, multiple questions for each type, and provides the correct and wrong answers.</figcaption></figure><p>The new version of QGen went live on all platforms in <strong>June 2023</strong>. We’ve been actively monitoring the adoption of the feature by our authors and the first numbers are very encouraging: <strong>more than 3 out of 4 authors</strong> who have tested our suggested questions feature have incorporated them into their course!</p><h3>III) Learnings and next steps</h3><blockquote>“Getting the model right is only half the battle. Surprisingly, the client-side implementation is equally as challenging as training and running the models.” (<a href="https://blog.replit.com/ai">https://blog.replit.com/ai</a>)</blockquote><p>Building our first generative AI feature at 360Learning has been an <strong>enriching learning experience</strong> that involved a very <strong>cross-functional team</strong>: <strong>ML engineers</strong> regarding modeling, finetuning, and, yes, some prompt engineering; <strong>product designers and managers</strong> for designing the frontend workflow and UX; <strong>software engineers</strong> for architecting the backend communication between the platform and the service and finally <strong>DevOps engineers</strong> for running the service on Kubernetes. While it’s (now) commonly advised to start building a generative AI feature with an external and generic model <em>before</em> moving to run your own internal and custom models, we had to go the opposite way. One major advantage is that we had already designed and validated<strong> a frontend workflow</strong> that would serve our customers best, and a <strong>backend framework</strong> flexible enough to switch models and APIs.</p><p>Does that mean we won’t go back to <strong>custom models</strong>? Nothing is sure yet. We’ll keep collecting feedback from our users on this improved version of QGen and explore whether there’d be a need to have company-level specialized models.</p><p>And beyond QGen, we’re building <strong>very exciting new features</strong> for content creators, such as transforming internal static documents into courses, providing prompt templates, etc. Learn more about what we’re building here: <a href="https://360learning.com/product/artificial-intelligence/">https://360learning.com/product/artificial-intelligence/</a>.</p><p><em>Special thanks to the AI squad (Yuhua Li, Hélène Giraud, Ismaël Aliouat, Guillaume Vibert, Gerardo Zenobi, Ismail Ennassef, Benoit Letournel) as well as the DevOps team (Mourad Maatoug, Pierre Malherbe, Thomas Auffredou)!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a1823ec7feb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/building-an-ai-powered-question-generation-feature-at-360learning-before-and-after-chatgpt-5a1823ec7feb">Building an AI-powered Question Generation feature at 360Learning, before and after ChatGPT</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[360Learning’s Data Journey: Building A Resilient & Cost-Effective Pipeline]]></title>
            <link>https://medium.com/360learning-engineering/360learnings-data-journey-building-a-resilient-cost-effective-pipeline-4204b40be29a?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/4204b40be29a</guid>
            <category><![CDATA[event-hub]]></category>
            <category><![CDATA[azure-event-hub]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[amplitude]]></category>
            <category><![CDATA[data-pipeline]]></category>
            <dc:creator><![CDATA[Guillaume Vibert]]></dc:creator>
            <pubDate>Fri, 20 Jan 2023 11:35:49 GMT</pubDate>
            <atom:updated>2023-01-20T11:35:49.161Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A tunnel of data" src="https://cdn-images-1.medium.com/max/1024/1*zSFZX0ygLJfBF4O9vHhF5Q.png" /></figure><p><em>As we scale up at 360Learning, we’re on a quest to improve our product with great data. The problems get bigger, the guessing gets harder. Since the beginning of 2022, we have started to incorporate more sources of data in the heart and veins of our company. Staying true to our core value of </em><a href="https://360learning.com/blog/transparent-workplaces/"><em>transparency</em></a><em>, here’s a closer look behind-the-scenes at how we’ve done that.</em></p><p>Once upon a time, I worked in the healthcare industry. The product was top quality and the clients were buying the upgrades. But because of the amount of testing and validation required, and the physical time needed to build a new hardware, we typically had to wait two years between starting the project and having it in the hands of the doctor.</p><p>The devices weren’t connected, so we had no information on the usage of the software. Did we do a good job? Was this feature that we spent months building actually useful? Was there a point in this optimization we took weeks to do?</p><p>To this day, I still don’t know. There is nothing we can’t build in software, so the only important question is: Should we spend our money building it?</p><p>At 360Learning, our business is special among the other scale-ups: we have a huge functional base. We build a product that is used by startups, mid-market organizations, and enterprise companies for onboarding, sales training, customer training, and a host of other learning and development objectives. Writing the “right” feature when you have so many different clients is a challenge. It gets very hard, very fast.</p><p>And to make the right decision, we need data on the usage of our platform.</p><p>Thankfully, at 360Learning, we have that data. And our data pipeline is only getting stronger.</p><p>This is the story of how it’s come to be.</p><h3>How stronger privacy policies affected data collection</h3><p><em>Why Steve Jobs failed us.</em></p><p>At 360Learning, our main source of truth is Amplitude. Our Product Managers chat with customers during discovery meetings to find out what needs they have, and based on their feedback, we create new features. Then, we implement and release the features with Amplitude trackers that monitor their usage.</p><p>It’s quite easy to see how often a button is used or how many times a page is opened, as well as analyzing the workflow of a user to see when they drop out due to one of many possible reasons: loss of interest, confusing user interface, unclear copy, etc..</p><p>Amplitude is a great tool, and it’s designed very well. It is easy to integrate inside your front-end code and to trigger a call from the browser to the Amplitude servers each time a user triggers a widget. However, this behavior works exactly the same way as the ad trackers that populate the internet ecosystem to track your every move and sell your data to the world.</p><p>A few years ago, Apple decided to make more billions by becoming a “champion of privacy” and started to wage a war against those trackers. Mozilla, Microsoft, even the kings of ad, Google, were all forced to follow and all the browsers (and adBlockers) are now blocking all or part of those trackers, by default or optionally.</p><p>Amplitude, as a somewhat innocent bystander, suffers the same fate as the rest of the trackers. Some events will pass through, depending on the version of the browser or the configuration of the network, but some won’t. At the end of the day, all of the data is not 100% available, and therefore not reliable. At best, it can be used to gather trends in usage (by supposing that the amount of data blocked stays the same across time and features).</p><p>Any tentative fix inside the browser is a losing battle — you’re stuck between the biggest companies of the world and their money.</p><h3>Sourcing more data outside of Amplitude</h3><p><em>Like chocolate, there is never too much data.</em></p><p>But there is a way to correct Amplitude’s data. It requires going server to server. Instead of having the browser send the data to Amplitude, we can push the information to our servers, and then do the call ourselves to Amplitude.</p><p>We <em>could</em> do that. That would mean some implementation work, some infrastructure and maintenance cost, maybe losing some quality of data (e.g., information on the browser used by the user), and definitely losing part of the appeal of Amplitude — the ease of use and simplicity and instead using it as an expensive database with a very nice user interface.</p><p>But we knew we wanted more, a better event management system. Our goal is not just tracking the usage of our widgets but to also answer :</p><ul><li>What is the usage of this feature per company? Per company type?</li><li>How much is the cost of maintaining this feature? Is it worth it?</li><li>Can we improve our Machine Learning models?</li><li>Can we reduce the usage of our legacy database? Can we speed it up?</li><li>Can we use Event Driven Architecture ?</li></ul><p>We want to show the value 360Learning brings to our clients, and we decided that fixing Amplitude wouldn’t be worth the investment, and we could go further: Track the data, get as much of it as possible, store it, and use it.</p><h3>How did we plug the data in?</h3><p><em>Let’s get technical.</em></p><p>First, we needed to gather the data. Each feature should be able to publish an event describing the change of state that happened, and it should be stored somewhere for further analysis. We would need to be able to quickly create events, push them at a high frequency (thousands per minute), and store them for at least a year. Bonus points if the technology used could also be used later as an input in streaming architectures.</p><p>We just finished a migration to the Microsoft Azure Cloud, and our DevOps squad still has opportunity to grow so building our own Kafka or HDFS cluster did not seem like the best use of our time.</p><p>Instead, we dipped into the managed technologies Microsoft proposed.</p><h3>Azure ServiceBus and Azure Logic Apps</h3><p>Azure ServiceBus is a message broker with publish/subscribe capabilities. We’re starting to use it as an inter-service protocol between our monolith and some new services. It was easy to set up, it promised to handle 100ks messages/minute, with a separation of concern by topics and a life expectancy of several days for the messages. All in all, something that sounded like everything Kafka can do, and well inside our requirements. It worked as promised, and we could publish our events.</p><p>To store our data, one candidate was to reuse MongoDB. However, we eventually decided against it, as we did not want to impact our current cluster. Plus, MongoDB is optimized for fast storage and fast reads, built on RAM. We preferred to go with Azure Blob, a cold storage (HDD), optimized to store massive amounts of data, where the Gb is so affordable you never think about deleting it.</p><p>However, there was no automatic feature to store the events on the ServiceBus. The point was to require minimal maintenance for the pipeline, so we did not want to implement a service — from scratch — that would consume and store the data.</p><p>So, we turned to the Azure Logic Apps, a “NoCode” functionality that allows us to build workflows (signals to actions). From a signal “Event in the ServiceBus”, react with “Write in Azure Blob”. However, the edge cases kept piling up and complicated the logic app: different types of blob storage, steps to decrypt the messages, file name format depending on the event type and date, etc…</p><p>In the end, we had a functioning pipeline but already an overly complex logic app.</p><figure><img alt="A diagram of logic apps with more than 15 boxes linked" src="https://cdn-images-1.medium.com/max/1024/0*spdSGmGbWZdQpP37" /><figcaption>It got too complex way too quickly</figcaption></figure><h3>Azure Event Hub</h3><p>We then decided to dig into the alternatives that Azure was proposing (there are at least 3 different ways of doing pipeline of messages). We realized that our workflow corresponds to Azure Event Hub much more, because for all intents and purposes, Azure Event Hub seems like a reimplementation of Kafka.</p><p>All basic concepts are there (producer/consumer, topics, retention, partitions, scalability), and as far as I saw, only one additional feature (you can have namespaces of topics on Event Hub, a way to gather your topics together). The key differentiator with ServiceBus was this one-toggle feature to “Capture”, i.e. store the events automatically in a Blob Storage. It was as simple as advertised and we could enjoy our events stored by a line of config, <em>and</em> we could dump the logic apps.</p><p>At this stage, our platform could emit events and they would be stored forever.</p><p>Now what about actually doing something with the data?</p><h3>Databricks</h3><p>Databricks is a software development platform used by our Machine Learning engineers. It allows them to write down Python code, read our MongoDB, process the code by providing on-demand servers, and write down the result in Cosmos DB (fully compatible with our MongoDB). We tested and validated that we could access the Blob storage as well.</p><p>We took a minute there and checked our goals. It was now possible for us to provide more data to the ML Engineers. But there were additional clients: we want to empower our dev with the capability to develop streaming applications and to be able to precompute a stack of data. Python and Spark/Spark Stream on Databricks is a good response, but the gap versus our current tech stack (Javascript and MongoDB) would make the adoption tough.</p><p>Aside from the pure tech discussions, we realized the pull for data was also strong in the Operations part of the company. How much is this feature worth to a client? And, subsequently, to us? In other words, how can we link the usage of the product to both their and <em>our</em> revenues?</p><p>Finding this information requires us to know about all the databases and sources of data of the company, how to link them, clean up the data, and realize its full potential.</p><p>Luckily, we had someone who knew how to do that.</p><h3>Building a Data Squad to support the data pipeline</h3><p><em>People come first.</em></p><p>Enter Julie. She’s our Data engineer and, at that time, she was working full time for Operations, answering the previous questions about businesses and usage of the platform as much as she could. But she grew frustrated with the information that was lost forever, the difficulties of getting all the data together, and the inability to provide the requested product data.</p><p>There were a lot of discussions, some time for reflection, and ultimately we realized that we needed more from her — we needed her skills to build a proper data pipeline, we needed her knowledge to be applied further in the company, and we realized that she would need some help to answer our new needs along with the old ones. So, we opened up some positions! And now, we’re looking to build out our ever-growing data team.</p><p>If all goes well, we envision the Data Squad of being composed of several Data engineers who will focus on the data pipeline, as well as Data Analysts, who will focus on harnessing this data to improve our product and business. The Data Squad will be incubated inside the AI Squad to start — as the ML engineers are huge consumers of data — until it gets big enough to be its own independent squad.</p><h3>What lies ahead for the team</h3><p>At the time of this writing, we have two event hubs plugged in prod. We have identified three other use cases (group membership, mail events, and subtitles used) that should be implemented during the quarter. Our data pipeline is getting stronger, with new tools in various stages of POC.</p><p>Hopefully, we will get back to you with status updates soon! And if you’re interested in joining the team, check out our <a href="https://jobs.lever.co/360learning?department=R%26D&amp;team=Engineering">open roles here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4204b40be29a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/360learnings-data-journey-building-a-resilient-cost-effective-pipeline-4204b40be29a">360Learning’s Data Journey: Building A Resilient &amp; Cost-Effective Pipeline</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[‘The Only Time Learning Was More Fun than Hacking the LMS’: My First 100 Days as a 360Learning…]]></title>
            <link>https://medium.com/360learning-engineering/the-only-time-learning-was-more-fun-than-hacking-the-lms-my-first-100-days-as-a-360learning-a56d90463e9f?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/a56d90463e9f</guid>
            <category><![CDATA[startup-life]]></category>
            <category><![CDATA[scaleup]]></category>
            <category><![CDATA[software-architect]]></category>
            <dc:creator><![CDATA[Guillaume Vibert]]></dc:creator>
            <pubDate>Mon, 31 Oct 2022 13:55:25 GMT</pubDate>
            <atom:updated>2022-10-31T13:55:25.646Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iotgYqpA047wvbNZWuuzRg.jpeg" /><figcaption>100 days is a short time</figcaption></figure><h3>‘The Only Time Learning Was More Fun than Hacking the LMS’: My First 100 Days as a 360Learning Architect</h3><p>I started working 20+ years ago, in the days of the Franc and the <a href="https://en.wikipedia.org/wiki/Wireless_Application_Protocol">WAP</a>. Since then, I’ve worked as a software dev, individual contributor, and tech lead in the healthcare industry, adtech, primitive web games, defense, and other sectors. I’ve worked on mobile (pre-smartphone), embedded software, and backend applications.</p><p>I’ve built a lot of things, including some resilient software that could last weeks without any issue, and some other that scaled up to million QPS, and some awful bugs in prod. But I never had the opportunity to be an architect, until today.</p><h3>How and why did you join 360Learning?</h3><p>I got pinged on LinkedIn by one of 360Learning’s tenacious TAs when I was feeling very down about my then job. She got me over my depression by showing me the opportunity to work on something fun and useful. I passed a few tests, had some great talks with a handful of people in the company and jumped.</p><p>I really liked the idea behind the product, and behind <a href="https://360learning.com/collaborative-learning/">collaborative learning</a>. I’d been through some awful online training in previous roles. I remember spending hours hunched above my computer, dying of boredom clicking “Next” to finish my mandatory training. At some point, I grew a market selling the answers of the final quiz for snacks. It was more fun hacking the tool than passing the courses, so it’s fair to say I found our platform refreshing.</p><p>The company is at an exciting moment. We are small enough that you can make a real impact every day, yet equipped enough to scale up to Unicorn status and more. The position was also intriguing. I clearly stated in my application that I had no previous experience in Javascript, which for most companies would have been a problem. At 360Learning it was accepted by everyone, which was another green flag that the company looked beyond the number of years working in a certain language, and trusted personal growth enough to let people develop new competencies.</p><h3>What did you expect from the job? Did you have any reservations?</h3><p>I’d never held the position of Architect before, and I had reservations. Most of this comes down to the way I’d seen Architects work in other roles. Often, Architects work separately from the dev teams, and climb down from their Mountain of Knowledge once a quarter to deliver the Answer, before climbing back and never being seen again for three months. I never liked this style of working, as it couldn’t correspond to the real problems we were seeing out in the field.</p><p>I’ve also seen architects as Saviors. Each time we encounter an issue, we would have to wait for them to come with the solution in a very top-down approach. This is also an unrealistic expectation on part of the Architect, and incredibly demeaning towards the skills of the squad.</p><p>Instead, I wanted to work collaboratively with a team to solve challenging problems and bring value to our clients. And talking to people at 360Learning before I joined, it was clear I’d have the chance to do this.</p><h3>What happened during your first 100 days?</h3><p>As we’ve talked about, <a href="https://360learning.com/blog/360learner-journey/">onboarding at 360Learning</a> is a very complete and comprehensive checklist. As it is part of the product that we are selling, it makes sense that it is so refined in our company.</p><p>When I joined, I had a board of tasks already prepared for two to three months. It was a bit overwhelming, but curiously reassuring. I knew exactly what was expected of me day after day, week after week until I had the time to find my feet. There was no self-questioning phase that could bring discomfort.</p><p>Concretely, I had the same ramp-up as a regular dev, which was crucial. I had a lot of training on the company, the tech, and the philosophy (Convexity). I started with simple, concrete tasks, then grew from basic debug, front-only, to back-only, and then full stack features. After a couple of months, I was comfortable with the codebase and the daily work.</p><p>Each product squad in 360Learning is assigned a referent Architect. After six to eight weeks, we dispatched the squads among the Architect team. It was a very gentle conversation where personal flavor was put first, with some tweaks like keeping squads that worked together with the same Architect.</p><p>In the end, I was assigned AI and Use Cases. Both squads were starting up and mostly composed of newcomers. That allowed us to ramp up together and find our favorite way of working. I joined the team, studied the tech stack and the product targets, and quickly jumped into the definition of OKRs for Q2. In parallel, I found out that we were missing a full data storage solution and started to study different solutions around it as the AI Squad had plenty of plans around that.</p><h3>How would you define your job now? What’s different?</h3><p>Right now, I’m working with three teams: my two product squads and the Architecture team. I’m trying to find an open subject on each team and split my time working on those. Maybe I should focus more!</p><p>My personal motto as an Architect is: “Help when things get tough”. I haven’t got enough time to do everything, so I need to identify my strengths and get the greatest value out of my time. That’s why it’s so important to take the time to discuss with the squads, challenge the <a href="https://www.productboard.com/glossary/epics/#:~:text=An%20epic%20is%20a%20large,start%20creating%20functionality%20from%20them">epics</a> and stories–but it means I can’t code as much. I try to have one ticket for coding opened up at every milestone.</p><p>We do cross-team code reviews at 360Learning, so I spent quite some time reading code on Github.</p><p>I’m also keeping some time to learn, I consume a lot of papers and videos about our current and potential tech stacks. I’ve got the experience and the taste of working with big data (and streaming application / Lambda architectures), and through talking about what the other 360Learners were doing (and not) I’ve found some place where I could do both and started building a data pipeline. Hopefully I’ll have more to say about that next time.</p><h3>Have you experienced any disappointments so far? Do you have any warnings for other candidates?</h3><p>As great as my experience has been at 360Learning so far, there are definitely some things we can improve on. I have a sense that our Architects are working too much in isolation from each other. We know we have to spend more time together, so we’re sharing more and more. Our new PM is helping us push in that direction.</p><p>This is an organization that gives you a lot of freedom. You shouldn’t expect people to come see you and tell you what you have to do. We expect our Architects to be mature enough to help us identify where we need to spend time and resources. This doesn’t mean working alone, but working collaboratively by asking the right questions, investigating the issue and learning &amp; teaching as much as possible.</p><p><em>Related: </em><a href="https://360learning.com/blog/prep-interview/"><em>So, You Want to Work at 360Learning? Here’s How to Prep for Your Interview</em></a></p><h3>What do you hope to achieve in another 100 days? 1,000 days?</h3><p>In a hundred days, I hope we’ve cracked open our monolith architecture and begin releasing services dedicated to smaller functions. We have a pipeline streaming tons of events per second and jobs consuming them for good reasons and great results.</p><p>In a thousand days, I hope we have services everywhere, with a DevOps mentality that allow us to frequently release new, performant features with a top level of quality. With our great product, we will have conquered the US, and we’ll be targeting even more challenging markets in the east and the south. Our R&amp;D team will be known across the world as a great place to work, fun and full of opportunities. Also, our IPO will have been a tremendous success!</p><p>If you’re interested in what the professional path for developers looks like at 360Learning, take a look <a href="https://360learning.com/blog/360learner-journey-engineer-career-path/">here</a>. And if you’d like to join the 360Learning team, check out our open roles <a href="https://jobs.lever.co/360learning">here</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a56d90463e9f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/the-only-time-learning-was-more-fun-than-hacking-the-lms-my-first-100-days-as-a-360learning-a56d90463e9f">‘The Only Time Learning Was More Fun than Hacking the LMS’: My First 100 Days as a 360Learning…</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[APIs, Squads, and More: Your Guide to Life As a 360Learning Software Architect]]></title>
            <link>https://medium.com/360learning-engineering/apis-squads-and-more-your-guide-to-life-as-a-360learning-software-architect-17902583c455?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/17902583c455</guid>
            <category><![CDATA[microservice-architecture]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[software-architect]]></category>
            <category><![CDATA[software-craftsmanship]]></category>
            <dc:creator><![CDATA[Thierry Lalande]]></dc:creator>
            <pubDate>Mon, 31 Oct 2022 06:38:16 GMT</pubDate>
            <atom:updated>2022-10-31T06:38:16.788Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eQegSBEmirr9e1zbWUpCnw.png" /></figure><p>On my business cards, my title says: Software Architect, 360Learning. Sounds cool, right? But what does it mean in reality?</p><p>In our previous article, my fellow Software Architect Guillaume told you all about his <a href="https://360learning.com/blog/360learner-stories-software-architect/">first 100 days at 360Learning</a>, offering an insight into how we do things around here — and some of the biggest surprises in store for new joiners.</p><p>In this article, I’m going to continue this tour of what it’s like to be a Software Architect at 360Learning by diving into my most important responsibilities and showing you how our team helps bring great learning support to our clients.</p><p>Of course, this is just a sneak peak of what is <em>behind the scenes</em>. But if you’ve ever wondered about joining our team, I’m here to give you an idea of what we’re working on at the moment, the “hot topics” for our fellow Software Architects, and what the future has in store for us!</p><p>Let’s get started with the basics.</p><h3>API (Application programming interface)</h3><p>You must have heard of APIs! They are everywhere, and most of them are designed as REST (REpresentational State Transfer), meaning they are basically ways to communicate between components, by sending self-described messages and waiting for a response which gives insight on whether it worked as expected.</p><p>One purpose of APIs is to hide the actual processes done to perform an action, exposing only the parameters required, and the result.</p><p>It’s our responsibility as developers and architects to make sure that:</p><ul><li>APIs visibility matches the need</li><li>No unnecessary API is created</li><li>No redundant API is created</li><li>Only the minimum amount of parameters is needed (no superfluous parameters)</li></ul><p>At 360Learning, our public API lets you automate several instructions related to users, groups, courses, and programs. By using our set of secured public functions, you can easily create connections between 360Learning and a third party software or a website.</p><p>In addition to this public API, the software architects team works at designing internal API as microservices, which helps ensure scalability along with asynchronous workers.</p><h3>CI/CD (Continuous Integration / Continuous Deployment)</h3><p>CI/CD pipelines are aiming at integrating changes in the existing codebase while ensuring that it does not break anything, and deploying this new code as soon as it’s ready, without waiting for a planned release.</p><p>When choosing the tools and steps within the CI/CD pipelines, you need to consider the following factors and constraints:</p><ul><li>Type of architecture (monolithic, service-oriented).</li><li>Kind of repository structure (mono-repository vs multi-repository).</li><li>Budget allocated to these tools.</li><li>Integration with version control system and infrastructure.</li></ul><p>When properly defined, a CI/CD pipeline allows developers to push changes several times a day without waiting for a planned milestone.</p><p>At 360Learning, we are running a CI each time a development branch is about to be merged in the main branch. Features are released during milestones every three weeks to avoid changing the platform too often, which could be disturbing for users.</p><p>We are working on integrating a CD tooling to bring flexibility to this process and get to a point where each developer can release as soon as a change is ready. Having automated steps will allow us to go quicker.</p><h3>Cloud</h3><p>Cloud computing is often achieved using one of the three main cloud services providers — namely Amazon Web Services, Microsoft Azure and Google Cloud. They provide a bunch of services designed to match the vast majority of use-cases.</p><p>When using cloud services, you can easily customize the number of servers you’ll use, their memory and CPU, use either dedicated servers or go to serverless architecture. Resources can be shared by several customers, lowering the global maintenance costs when compared to on premise infrastructure while improving security and reliability.</p><p>At 360Learning, we moved our infrastructure earlier this year from our on-premise infrastructure to Microsoft Azure. Now we can scale more quickly and efficiently the servers we use or even move a bit further to serverless.</p><h3>Data</h3><p>Data is core for every company, and architects have to structure it well. For example, well-designed databases help make everything go smoothly, but databases designed without thinking about future growth in data volume may lead to slow or even failing platforms and services.</p><p>Here’s what we consider when deciding our data storage options:</p><ul><li>Do you need to store temporary key-values or counters? Then you may need an in-memory datastore (like Redis, Memcache).</li><li>Do you need to store data you don’t need to access often and/or fast? Then a cold storage is a choice you may consider (like HDFS, DataLake).</li><li>Do you need to store persistent and well-structured data, while accessing it fast? Then a hot storage is what you need (like MongoDB, SQL).</li></ul><p>At 360Learning, our platform is historically built around MongoDB, where we store all of our data. This has led to some difficulties as the amount of data stored this way is growing quite fast.</p><p>That’s why the architect squad is evaluating alternative solutions like Redis (to store temporary key/value data), and Snowflake (for Cold Storage). Introducing a data lake and a data warehouse will help us address our AI / BI / Dashboard use cases.</p><p>In fact, we’re so into data at 360Learning that we’re building a Data Team! More on that later.</p><h3>Monolith vs Microservice Architecture</h3><p>In a monolithic architecture, everything is in one place. This can from time to time lead to some ‘spaghetti plate’ architecture.</p><p>In a microservice architecture, each component is an independent piece of software with inputs and outputs clearly defined. The idea behind such an architecture is to provide ‘black box’ services doing their stuff without the other services knowing — or even needing to know — how it works.</p><p>Microservices have some important advantages:</p><ul><li>Improved elasticity</li><li>Flexibility on technology choices (language, storage, hardware)</li><li>Faster delivery</li><li>Rolling update</li><li>Resilience</li></ul><p>The 360Learning code base has historically been a big monolith. This helped us quickly deliver features and adapt to our clients.</p><p>However, it has now reached a point where we need to cut it in pieces to keep this agility and performance. That’s why building independent services with a clearly defined scope will help us enforce modularity and ship faster.</p><p>You can expect more news about that in the upcoming year.</p><h3>Referent</h3><p>At 360Learning, each squad has its own referent architect; and each architect acts as a referent for several squads. What does it mean exactly? Developers in squads are working in a very specific part of the codebase, and may not need to interact with other squads except in some particular cases.</p><p>The referent architect helps to bring a more global vision in cooperation with other architects to make sure everything will work properly when interacting with other parts of the code. The referent architect is also here to help design complex features or data structures, to make sure it will behave as expected while the volume of data will grow.</p><p>Sometimes, a feature can look simple at first sight, but needs deeper analysis to check if the volume will not break the ability to load data, or to make sure it will still work when traffic starts to grow. That’s why our reference architects play such a key role.</p><h3>Repository</h3><p>When working on a codebase shared among developers, you need to choose how code is split — or not — and how everyone can work together without blocking each other. Basically, it’s about choosing either a monorepo (everything is in the same place) or multirepo (the codebase is split, e.g. in services, and each service is on its own repository).</p><p>There is no clear winning option here. While we’re using a monorepo for now, we are starting this conversation at 360Learning, especially in the context of moving from a monolith to microservices. It already sparked some fiery debates! More is coming soon.</p><h3>Security</h3><p>As for every company, security is key for 360Learning. It’s a key part of the role of architect to make sure features are securely designed regarding authentication and visibility as well as data integrity.</p><p>That’s why we take part in security audits requested by customers to bring a global 360Learning codebase vision. 360Learning has been recognized as one of the cybersecurity champions by Cyrating. Click <a href="https://www.linkedin.com/posts/cyrating_cybersecurity-frenchtech-next40-activity-6940192559105257472-ouZ3/">here</a> to learn more.</p><h3>Squads</h3><p>Teams, tribes, communities, chapters, guilds, squads: in every company, splitting developers — and other types of roles — is necessary to work efficiently on various projects. Each group works as a whole to deliver features.</p><p>At 360Learning, developers are dispatched among squads. Each squad focuses on a specific functional scope (e.g. Authoring, Dashboard, Stickiness).</p><p>Members of a squad are:</p><ul><li>A Product Manager</li><li>A Product Designer</li><li>An architect referent</li><li>One or more developers</li></ul><h3>Software craftsmanship</h3><p>Developers need to think about code quality — computational complexity, readability, maintainability — when they implement a feature.</p><p>Architects have a role to play here, as we tend to have more experience working with a range of different coding practices and technologies…It’s all about finding the right balance between a code easy to understand and maintain, and a very efficient code (sometimes, you just can’t have both).</p><p><a href="https://www.amazon.com/Clean-Code-Handbook-Software-Craftsmanship/dp/0132350882">Robert C. Martin’s Clean Code handbook</a> is mandatory reading at 360Learning, and we can only encourage you to read this book!</p><h3>TypeScript</h3><p>Historically, our whole codebase was in Javascript. We decided to migrate to TypeScript in order to help keep the code clean.</p><p>Now a large part of the backend code is in TypeScript, and every new code is written using this language. Of course it adds some strong constraints — that’s the goal of a typed language — but it also helps increase robustness.</p><h3>That’s all for now</h3><p>By following along through these few topics, I hope you have now a taste of how Software Architects at 360Learning work together with developers, product managers and more generally every stakeholder.</p><p>From Micro-services to data management, these are just some examples of what we are working on in the 360Learning Software Architect squad. But there is more! I could have written a whole article just on Event-Driven architecture, or how we’re using AI and Machine Learning. There’s more to come soon on these topics, so stay tuned!</p><p>And if this tour of life as a Software Architect has piqued your interest in joining the team, great news: <a href="https://jobs.lever.co/360learning/d6797783-c116-4687-8ee2-900c4ebcd29d">we are hiring</a>! Feel free to get in touch with a member of the 360Learning team to find out more.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=17902583c455" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/apis-squads-and-more-your-guide-to-life-as-a-360learning-software-architect-17902583c455">APIs, Squads, and More: Your Guide to Life As a 360Learning Software Architect</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Key takeaways from our Kotlin Multiplatform journey]]></title>
            <link>https://medium.com/360learning-engineering/key-takeaways-from-our-kotlin-multiplatform-journey-764415e3606f?source=rss----79e2939dd800---4</link>
            <guid isPermaLink="false">https://medium.com/p/764415e3606f</guid>
            <category><![CDATA[kotlin]]></category>
            <category><![CDATA[ios]]></category>
            <category><![CDATA[kotlin-multiplatform]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[mobile]]></category>
            <dc:creator><![CDATA[Yev Kanivets]]></dc:creator>
            <pubDate>Wed, 11 May 2022 07:08:23 GMT</pubDate>
            <atom:updated>2022-05-11T07:08:23.769Z</atom:updated>
            <content:encoded><![CDATA[<p>Kotlin Multiplatform is one of the hottest topics in native mobile development this year. KMP gives us an unprecedented opportunity to share business logic written in Kotlin between Android and iOS native mobile applications.</p><p>The technology has already proven itself a production-ready solution applied by many large, medium, and small companies, even though it’s only in the alpha stage.</p><p>The developer community is super excited about KMP, with dozens (or even hundreds) of articles and sample projects published monthly. Now, it’s almost certain that developing your next native mobile app using KMP is a great idea.</p><p>But what’s about existing projects? Can we use KMP in legacy mobile apps developed by separate Android and iOS teams? What are the cons and pros?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p2WTZJa1t3liiJDBmkUbFg.png" /></figure><p>In this article, I’ll share key takeaways from using Kotlin Multiplatform in 360Learning native mobile apps for almost two years to improve consistency, reduce the number of bugs, and finally, increase dev productivity.</p><h3>Context</h3><p>I’ve previously shared our journey to adopting Kotlin Multiplatform at 360Learning in <a href="https://medium.com/360learning-engineering/the-pursuit-of-consistency-on-mobile-c129d8a9dd96">The Pursuit of Consistency on Mobile</a>. It includes all the necessary context, leading us to the first feature done with KMP.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u6dOzD3k6GnZtJ36SpHxeg.png" /><figcaption>Our GitHub activity</figcaption></figure><p>In short, we are a team of seven mobile developers, one designer, and one product manager working on native mobile apps for Android and iOS. There are 25K+ commits, with the first contributions made in 2015.</p><p>Most of the developers in our team do both Android and iOS native development, so Kotlin Multiplatform came very naturally to us.</p><h3>Source code structure</h3><p>Kotlin Multiplatform enables sharing of the code between different platforms — iOS and Android, in our case. But where should this code be located?</p><p>As for most teams working on native mobile applications, it’s natural to write common (KMP) code as a part of the Android application (both are in Kotlin), hence in the Android app’s GitHub repository. iOS repository hosts iOS-only source code with a script to sync and build common Kotlin code.</p><p>This approach has several significant issues:</p><ul><li>Every time we need to change common Kotlin code from the iOS side, we make that change in the Android repo, make a commit, then Pull Request, have it reviewed, merged, and only then does it become available on iOS.</li><li>All changes made in the common code from the Android side aren’t synced to the iOS side right away because we depend on a specific commit. It creates an alignment debt, which increases with time and is paid off by the first “lucky” person who needs an update of the common code from the iOS side.</li><li>At the start, the iOS app depended on the latest version of the Android’s dev branch. It allowed fetching common code changes right away. But it was breaking the iOS app unexpectedly producing a horrible dev experience, so we decided to depend on a specific commit instead.</li></ul><p>This is why we’ve decided to migrate to the mobile monorepo, which hosts both applications. The great bonus is that now we can structure the whole source code by feature modules that contain common, Android and iOS code.</p><p>The third option preferred by (much) larger teams is having common Kotlin code in libraries written separately from both applications. Needless to say that changing common code from both sides is complicated.</p><h3>Project structure</h3><p>Modularization seems to be the industry’s standard for structuring from medium to large mobile applications. Such structure ensures proper separation of concerns, encapsulation, and scales along with a team.</p><p>At 360Learning, we have about a hundred modules total in both applications, with 10+ modules written to be shared between iOS and Android using Kotlin Multiplatform. Sounds good so far?</p><p>After adding the third or maybe fourth module, we discovered that while multiple KMP modules are working expectedly well on Android, they don’t work on iOS. At least, they don’t work the way everybody expects.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Rr7blak68PMmAy0E.png" /></figure><p>I’ve described this issue in the <a href="https://medium.com/xorum-io/three-framework-problem-with-kotlin-multiplatform-mobile-16267c5afa53">Three-framework problem with Kotlin Multiplatform Mobile</a> in great detail. But in a nutshell, Kotlin/Native compiles every module and its dependencies into a single iOS framework to be included in the iOS application.</p><p>It means that all dependencies are duplicated for every module. So such simple things as an authentication module (to share the bearer token) don’t work because each module contains its own copy of it. To my knowledge, JetBrains doesn’t plan to fix this problem.</p><p>The recommended solution, which we use, is to create an umbrella module that depends on all KMP modules and include it as a single massive framework in the iOS application. It doesn’t affect the Android codebase. The proper architecture ensures no misuse of globally available APIs on iOS.</p><h3>API structure</h3><p>Kotlin Multiplatform isn’t a cross-platform framework but a multi-platform code-sharing solution. It means that we, as developers, have great flexibility to choose what and how to share between different platforms.</p><p>With great flexibility comes great responsibility. At 360Learning, we use standard Model-View-Presenter architecture with Navigators, Interactors (aka Use Cases), Repositories, Data Sources — powered by Clean Architecture.</p><p>For all new features, we apply Kotlin Multiplatform from the beginning. In our case, it means sharing everything but View. Yes, only the View is implemented in a platform-specific way on Android and iOS.</p><p>The View layer only interacts with the Presenter level and needs to know its interface without any implementation details. Hence, in the KMP modules, we declare Contracts that contain View, Presenter, Navigator, and UiModel.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/56291f70fc45a377e300edeec5bc4e94/href">https://medium.com/media/56291f70fc45a377e300edeec5bc4e94/href</a></iframe><p>So only the Contract layer is visible from the outside of the common module by default. Everything else is internal, even though we sometimes need to make exceptions with legacy codebases like ours.</p><p>By the way, any non-internal class, method, or variable in the KMP module is automatically exposed in the Objective-C header of the iOS framework compiled by Kotlin/Native. It clutters module API and increases app size.</p><h3>Development Flow</h3><p>We use GitHub to manage our work (projects, milestones, tickets). Since we had two separate repositories for Android and iOS apps, we also had separate tickets, and they were supposed to be done independently and shipped at the same time.</p><p>The issue with a common Kotlin code is that it needs to be written only once, usually while implementing the Android ticket. Hence iOS tickets depend on Android — something new for most mobile teams. So it requires more and better communication between devs from different platforms.</p><p>In our case, the two tickets of most of the features were done sequentially by the same people on both OSs, which is more effective in person-hours but takes longer to ship the feature since the work can’t be parallelized. iOS was always lagging behind.</p><p>With the migration to monorepo, we also decided to have a single ticket for both OSs, so they are implemented, reviewed, and shipped simultaneously (if done by the same dev). We use labels to mark platform-specific issues or bugs if needed.</p><h3>Developer Experience</h3><p>No technology can succeed if it offers an inferior developer experience (hello Xamarin). It includes everything from SDK and IDE to the standard library and community.</p><p>Our experience with Kotlin Multiplatform is built around native mobile IDEs such as Android Studio and XCode / AppCode.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DV88pES3l9cw43sg.jpg" /></figure><p>We aren’t using KMM plugin for Android Studio due to two reasons:</p><ul><li>it didn’t exist when we started with KMP</li><li>it doesn’t provide many benefits with two separate repositories</li></ul><p>As mentioned above, a common KMP module is usually implemented while working on the Android ticket without much struggle. Then the integration is done on the iOS side, which is generally the most complicated part.</p><p>Our first attempted features required a lot of changes in the common code to make it work on iOS. Mainly due to the technical limitations described in the following section. Some syntax of Kotlin just can’t be translated into Swift.</p><p>We’ve found what’s working and what’s not during the last two years. Hence we limit the set of keywords/tools while writing the common KMP code, so it works out of the box on iOS without much struggle.</p><p>Still, it’s highly beneficial to the team to clearly communicate with iOS devs the contract and expected behavior of the common code. Plus making sure that initial integration is working (mainly Dependency Injection), and supporting them along the way (pair programming and debugging).</p><p>Debugging the common code from the iOS side is difficult with two separate repositories. In our case, it’s done by making changes in the Android repo, pushing it, and fetching it on the iOS side by commit hash. It takes way too long to debug an issue like that.</p><p>With a migration to monorepo, we will start using the KMM plugin to simplify the debugging and the overall developer experience. For example, we will include common KMP code in XCode directly.</p><p>Our goal is to make the iOS dev experience with Kotlin Multiplatform as great as Android’s one. There is still a long way to get there.</p><h3>Technical Limitations</h3><p>Kotlin Multiplatform is still in Alpha with Beta planned for this spring. We started to use this technology pre-Alpha and were aware of the risks. We didn’t use it for critical features to stay safe.</p><p>The list of technical limitations we’ve discovered is quite extensive and mainly related to Kotlin/Native interoperability with Objective-C/Swift. We highly recommend reading <a href="https://kotlinlang.org/docs/native-objc-interop.html">the complete doc</a> before starting your first KMP module.</p><p>Here are a few most annoying limitations we’ve faced so far:</p><ul><li><a href="https://dev.to/touchlab/practical-kotlin-native-concurrency-ac7">Multithreading and memory management model</a>, which is the main problem targeted by JetBrains this year. It’s already fixed in <a href="https://touchlab.co/kmm-memory-model-when-to-switch/">Kotlin 1.6.+</a>, but performance is to be improved. As a temporary workaround, we wrap all calls to mutable KMP objects to be executed on the UI thread.</li><li>Some Kotlin features, like inline or sealed classes, coroutines, and others, are not well supported in Obj-C/Swift. So we decided to not expose them in the common code APIs at all.</li><li>All primitive Kotlin types are wrapped in Kotlin... classes on the iOS side. It means that we need to convert them to and from Swift primitives.</li><li>We use interfaces for scoping other interfaces (to define a contract), which isn’t possible in Obj-C/Swift, so the resulting protocol (interface) names are just long strings and sometimes difficult to find and strange to use (e.x. MySessionsUiModel.ContentSession).</li><li>Matching method names from different interfaces, like setUiModel, clash even if they take different parameters. Kotlin/Native adds _ every time such a clash happens, which becomes annoying after a dozen cases (e.x. setUiModel(uiModel___ uiModel: SessionStatsContractUiModel)).</li><li>We are using Koin for the Dependency Injection in KMP, but the graph is only checked in runtime and only for those nodes we use. Combined with iOS lagging behind Android, it can be tricky to understand what’s to inject.</li><li>More low-level issues are described in <a href="https://medium.com/360learning-engineering/the-pursuit-of-consistency-on-mobile-c129d8a9dd96">this article</a></li></ul><p>As a team, we simply acknowledge these limitations and find workarounds or adapt our practices to not face them anymore. There are no real roadblocks.</p><h3>Conclusion</h3><p>If I had just one word to characterize our experience with Kotlin Multiplatform, that would be “flexibility”. If I had one more word, I would add “great”. And for the third word, I’d say “too great flexibility” :)</p><p>We had all the options on the table when we started with Kotlin Multiplatform. It took us two years to find our ways of doing things efficiently. And discard those options that don’t work, at least for us.</p><p>Now when the KMP has almost reached the Beta stage, it’s vital to lower barriers for entry, provide the default ways of doing things, share experience and case studies. This is one of the reasons we share our experience in this article. We hope it will be helpful for other developers and companies.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=764415e3606f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/360learning-engineering/key-takeaways-from-our-kotlin-multiplatform-journey-764415e3606f">Key takeaways from our Kotlin Multiplatform journey</a> was originally published in <a href="https://medium.com/360learning-engineering">360Learning Engineering</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>