<?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[Stories by Yana Arbuzova on Medium]]></title>
        <description><![CDATA[Stories by Yana Arbuzova on Medium]]></description>
        <link>https://medium.com/@yanaarbuzova_92529?source=rss-da4ff3d1388c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ZNkRk5ijpcoHcs4AhmFVkw.png</url>
            <title>Stories by Yana Arbuzova on Medium</title>
            <link>https://medium.com/@yanaarbuzova_92529?source=rss-da4ff3d1388c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 19:10:33 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@yanaarbuzova_92529/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Global Ledger, a Ukrainian Crypto ML Startup, Enters the Global Market]]></title>
            <link>https://medium.com/sigma-software/global-ledger-a-ukrainian-crypto-ml-startup-enters-the-global-market-f541f86c92c7?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/f541f86c92c7</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[cryptocurrency-investment]]></category>
            <category><![CDATA[crypto]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Tue, 16 Nov 2021 09:03:57 GMT</pubDate>
            <atom:updated>2021-11-16T09:03:57.681Z</atom:updated>
            <content:encoded><![CDATA[<p>A Ukrainian startup <a href="https://glprotocol.com/">Global Ledger</a> has recently expanded to the global market by signing a strategic partnership with Promethiem AG, a Swiss FinTech group.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E3pxhQ-pIzNs9ZWjzeYfug.png" /></figure><p>The startup was founded by Alex Fisun and Dimitrii Bilokon, and facilitated in IdeaSoft (member of Sigma Software Group) business incubator. The product offers a unique service that dramatically disrupts money laundering and cybercrime allowing banks to freely work with cryptocurrencies. This is the result of Sigma Software Group and IdeaSoft continuous strategy to find and facilitate product business ideas within the service business.</p><p>ABOUT THE PRODUCT</p><p>Many countries have already implemented legislation that allows banks and financial institutions to operate with cryptocurrencies. However, there are mandatory regulations, including Anti-money Laundering and Counter Terrorist Financing legislative package that banks must follow when working with cryptocurrencies. To comply with them, they have to use the products that allow them to check where the cryptocurrency came from and assess the risk score of working with a particular client.</p><p>The Global Ledger success story began three years ago. The founders had a deep background in various Blockchain and FinTech projects and clearly saw the future issues that would arise upon regulations of the industry towards mass adoption of the crypto. The project was implemented using IdeaSoft services.</p><blockquote>“We managed to create a product that solves those risks. It took us more than a year to build it due to the complexity of the product before getting our first clients. Since then we have worked with government agencies in several European countries, with banks in Switzerland, and attracted several large customers,” says Andrey Lazorenko, CEO and Co-Founder of IdeaSoft.</blockquote><p>Global Ledger is instantly available in urgent situations and has affordable subscription packages for companies of any size. Allowing users to perform investigations and verify their counterparties, it improves the cybercrime response time and increases laundering publicity. With a user-generated database, reports, and alerts, the tool provides best-in-class sources of funds verification with way better visibility than solutions currently available on the market.</p><p>BIG OPPORTUNITIES FOR BANKS</p><p>Global Ledger opens many opportunities for banks and crypto exchanges, including:</p><ul><li>Monitoring and preventing AML/CTF risks in real-time</li><li>Ability to detect and counter the latest money laundering technologies, including mixing, smart contracts, NFT fraud, and more</li><li>Performing extensive due diligence procedure on entities and individual accounts</li><li>Integration with your current AML/CTF processes</li><li>Collecting all needed information for Suspicious Activity reporting in seconds</li><li>Leading investigations on money laundering in crypto</li></ul><p>The main advantage of Global Ledger over the competitors is data privacy — the gathered data is not transferred anywhere else. All the data is stored on the client`s side, which allows banks and financial institutions to stay GDPR compliant.</p><blockquote>“When it comes to working with cryptocurrencies, the banks’ main fear is the lack of opportunity to check where the money comes from. Global Ledger solves this problem and opens the doors for the banks to finally get their piece of the huge cryptocurrency market,” underlines Hanna Khrystianovych, Fintech Program Manager at Sigma Software.</blockquote><p>ABOUT THE STRATEGIC PARTNERSHIP</p><p>Recently, Global Ledger has signed a strategic M&amp;A partnership with Promethiem Group with secured funding of 3.5 million USD and access to new markets.</p><p>“Our ultimate mission is to disrupt criminal activity using blockchain transparency and traceability. We are happy to see new partners that share the same values and goals. Together we can push crypto compliance technology to the point where financial institutions and law enforcement will be able to detect and respond to money laundering in a fraction of a second,” says Alex Fisun, CEO and Co-Founder of Global Ledger Protocol.</p><p>“For the Ukrainian market, this is the first and so far, the only case when a crypto ML product has entered the world market. We see great prospects in this partnership including the opening of Global Ledger offices in Europe and the expansion of cooperation with Swiss banks. More and more European countries (and Ukraine is among them) are introducing cryptocurrency legislation, which opens new opportunities for us around the world and in Ukraine,” says Andrey Lazorenko.</p><blockquote>“Global Ledger is a comprehensive product that Ukrainian and foreign banks are already highly interested in. This is an example of a successful case of facilitating a product creation within a service company, and I think it has every chance to take on the global market,” says Valery Krasovsky, CEO and Co-Founder of Sigma Software.</blockquote><p>“At Sigma Software we also had similar cases, when products desperately needed by the market were created and developed on the basis of a solution found in an IT service business. One of the vivid examples is Clean.io, which became the Startup of the Year in Baltimore in 2019. The solution prevents malicious advertising and is currently used by dozens of big publishers across the USA. Sigma Software engineers were co-authors of this unique solution. Having years of hands-on experience in the domain, they spotted the unsolved problem and came up with a solution that immediately gained high level interest from potential customers. Likewise, Global Ledger, giving a real boost to banks development, can eventually become a standard in the financial sector,” he adds.</p><p>Andrey Lazorenko notes:</p><p>“By the way, we are launching a new crypto product soon and looking for a rainmaker, who will overtake the CEO role of the next Blockchain unicorn.”</p><p>Apply at hello@ideasoft.io or write directly to lazorenko.a@ideasoft.io</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f541f86c92c7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/global-ledger-a-ukrainian-crypto-ml-startup-enters-the-global-market-f541f86c92c7">Global Ledger, a Ukrainian Crypto ML Startup, Enters the Global Market</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[3 Winning Factors for Successful Development of Startups]]></title>
            <link>https://medium.com/sigma-software/3-winning-factors-for-successful-development-of-startups-6c0f2e0cb120?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/6c0f2e0cb120</guid>
            <category><![CDATA[startups-in-israel]]></category>
            <category><![CDATA[outsourcing-for-startups]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Fri, 24 Jan 2020 10:40:16 GMT</pubDate>
            <atom:updated>2020-04-03T11:15:51.242Z</atom:updated>
            <content:encoded><![CDATA[<p>According to <a href="http://www.moyak.com/papers/business-startups-entrepreneurs.html">estimates</a>, there are over 472 million entrepreneurs worldwide with approximately 100 million new businesses opened each year around the world. Historically, only 10% of those businesses survive. But don’t be hasty to bury the idea of your own startup, data shows that the number of successful startup exits grows year after year — 4,228 startup exits occurred worldwide in 2018.</p><p>How do they do it? What are the key prerequisites for a startup success? To unearth the success factors, let’s first look at the most common reasons for new companies breakdowns.</p><p>As statistics show, there is quite a list of reasons named for why the things went wrong.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/955/1*e2B9VA8xP3gZm1LIFXl9FA.png" /><figcaption>See the full version: <a href="https://www.statista.com/chart/11690/the-top-reasons-startups-fail/">https://www.statista.com/chart/11690/the-top-reasons-startups-fail</a></figcaption></figure><p>After analyzing the recited reasons and comparing them with the observations we made through 15 years’ experience of custom software development for startups from MVP to successful exits, we come to an interesting conclusion. Many of these reasons stem from just three factors that ultimately define how a startup will go through paces.</p><h4>1. A startup Idea That Comes from Real Business</h4><p>The first factor of startup success relates to the core of each new business — its idea. Actually, most ideas for startups originate from real business. However, there is one important thing to consider. This idea needs to be checked, then cross-checked, and cross-checked again to make sure it is based on actual business need, not just the founder’s guess about it.</p><p>Several years ago, we had two startup clients, one from the US, another from Europe. Both in the automotive domain, both with ideas, which looked pretty feasible at first sight and solved a relatively similar problem in the business domain they knew pretty well. The major difference between those two startups was in their approach to building the product — while the first one rushed to get the product delivered as fast as possible, the other spent one more month talking to potential customers, reworking the product vision, and considerably adjusting it to actual business needs. Guess, which business turned out to be more successful? The one that took time to refine the idea.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dDkEVFuGvYlbSKqo.png" /></figure><p>You may say that delivering the product to market is important for a startup, and it’s true — time-to-market is crucial in the startup world. But in order to turn into a key to your success, the startup idea needs to be cross-checked at the very early stages by several players in the domain, be it potential customers, investors, or the market itself.</p><h4>2. The Startup Raises External Money</h4><p>This may come as a surprise for some of you, but most startups are founded by personal savings, friends and family (<a href="https://www.entrepreneur.com/article/230011">https://www.entrepreneur.com/article/230011(link is external)</a>). It does not make them much good though. Even if you have enough own funds, even if you can afford losing them, raising external funding is still important for making your startup a success. Why so? Let’s look a little deeper.</p><p>Two reasons why external money is essential:</p><ul><li><strong>Speed and More Time to Reach Profitability. </strong>Additional funding gives startup founders a chance to hire more professionals and have their idea brought to life quicker. Also, the investment “safety bag” makes your startup more durable and grants you the possibility to spend more time on finding the market fit.</li><li><strong>Additional possibilities for marketing and promotion</strong>. Quite often, startups focus on developing their solution and route almost all funding to those activities. Most common keywords startups use in their search (after funding, of course) would be about architecture, web development for startup, UI, technologies, app development for startups. What is left behind in this scenario? Promotion and marketing, which make an essential part of product success. External funding can help cover all necessary expenses for harmonious product development and promotion.</li><li><strong>Additional check of your idea and product viability</strong>. Finding organizations of angel investors, who trust you with their money, requires additional homework to be done in terms of pitching, shaping the idea, and envisioning your startup strategy, which itself is very important and valuable for your startup progress.</li></ul><p>If you still doubt whether raising external money is a good idea, here’s a <a href="https://pitchbook.com/news/articles/heres-why-the-biggest-unicorns-in-the-world-are-taking-on-billions-in-debt">story(link is external)</a> that may convince you. Large unicorn startups that know their onions — Uber, Didi, Airbnb — do that too. Though these companies make a lot of money, they continue attracting millions in debt to develop their business before their competitors do it.</p><h4>3. The Startup Team is Right in Terms of Composition and Motivation</h4><p>It is no secret that many investors consider the team as the crucial criterion, when evaluating a startup and deciding if they want to invest in it. But what stands behind the term “the right team” and what is the gold standard for a startup dream team? Of course, a head-by-head composition of a team depends on the nature of your startup idea and the product or service you try to infuse to the market. However, some building blocks of a functional startup team remain the same across industries and product types.</p><p>KEY PLAYERS OF A FUNCTIONAL STARTUP TEAM:</p><ul><li><strong>CEO.</strong> A good CEO is a very important figure for any startup. This should not necessarily be a startup founder. The founder provides a good idea, while a CEO must be the right professional with knowledge and experience in business administration, and understanding of your domain. A perfect candidate for this position has management experience, is passionate for new challenges, and burns with the idea. Besides that, this person should be 200% dedicated to the startup and ready to invest all their time into driving your company to success.</li><li><strong>Trendsetter</strong>. Though a trendsetter role is not often met in startup teams, having one in your team can make all the difference. Who is that person? The trendsetter knows the business domain and is known in the business domain. This person will explain and advocate your idea to the market, and inspire your own team. A perfect candidate for this position is a public person, well-known visionary with a number of valuable connections and acknowledged reputation in your business domain. A trendsetter for your startup needs to be followed and trusted by a substantial number of professionals, burn with your idea and be willing to share it with other people.</li><li><strong>Execution team</strong>. To make a team out of individual specialists, they need to be committed to a common idea and work in close contact with your CEO and each other. Your team may include various startup professionals, the essential part is that they must be ready to wear many hats. At initial stages, the team is usually not that big, and roles and responsibilities are not that well divided as in a mature company — things grow and change too quickly. The problem of growth and scalability is inherent to all startups, and that’s where a reliable tech partner comes into play.</li></ul><h4>OUTSOURCING SOFTWARE DEVELOPMENT FOR STARTUPS</h4><p>In a fast-growing tech startup, you cannot have all specialists that you may happen to need. Also, you can hardly scale your team up and down as quickly as the development process requires. All these circumstances make outsourcing software development to a reliable tech partner very beneficial for a startup. As <a href="https://www.entrepreneur.com/article/326492">some entrepreneurs point out, 74% of companies outsource IT(link is external)</a>. That is used by new and mature companies as a way to increase a company talent pool, get access to required technical and/or managerial expertise, and save on employment costs. Moreover, even the most <a href="https://www.pixelcrayons.com/blog/10-successful-startups-were-built-with-outsourced-development/">high-standing companies outsourced software development(link is external)</a>, including Google, IBM, GitHub, Opera, etc.</p><p>A good idea for an early-stage startup in search of their market fit and, correspondingly, with instantly changing focus and tasks is to <a href="https://sigma.software/landing/dedicated-software-development-team/#reliable-tech-partner">hire a dedicated team</a> (we discuss dedicated teams in detail in <a href="https://sigma.software/about/media/what-consider-hiring-dedicated-software-development-team">this article</a>). There are many software development companies for startups to choose from, but if put in a nutshell, just make sure that the supplier you chose has experience of working with startups, has a scalable approach, and a wide range of services you may need. The latter will save you the trouble of looking for dozens of suppliers for each type of work and communicating all of them together.</p><p>Having a supplier that knows how to work with startups and understands their needs grants more benefits. For example, they may have special startup-oriented services and offers, such as our <a href="https://sigma.software/services/startups-cto-service">CTO-as-a-Service</a>. So, it’s a good idea to start looking for a software development firm to support your startup among those having experience of working with such companies.</p><h4>SIGMA SOFTWARE INITIATING UKRAINIAN TECH DAYS IN ISRAEL</h4><p>Sigma Software is organizing Ukrainian Tech Days in terms of 2020 the <a href="https://summit.ourcrowd.com/">OurCrowd Global Investor Summit</a>(Jerusalem), one of the leading startup events in the world, on February 11–13. For one week in February, the startup world gathers in Jerusalem for the OurCrowd Global Investor Summit, one of the premier events in the tech industry, and the largest business conference in Israel. In 2019, over 18,000 people from 189 countries registered to attend what has become the fastest-growing tech conference in the world. Entrepreneurs, venture capitalists, corporate executives, investors, and others came to the Innovation Nation to meet, learn, do business — and discover the most historic city in the most exciting country in the world. As Forbes reported, Summit Week in Israel was “the place to be.”</p><p>Our team is always happy to meet new entrepreneurs and share our best practices of helping startups to grow faster. As well as to show the world tech potential of Ukraine, a region with 200 thousand tech specialists and 110+ R&amp;D centers of the leading tech enterprises, good IT service companies, promising startups, and product companies. This is why we initiated <a href="https://www.facebook.com/events/1831630260302617/">Ukraine Tech Days in Israel</a> at OurCrowd. Together with other prominent companies — Ciklum, Shalb, Beetroot, InDevLab — we will run the set of events dedicated to delivering high-level software development services for startups.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*13SzTPi0KZn3tXflQNnEng.png" /></figure><p>What’s included in Ukraine Tech Days? There are three days in Tel Aviv and in Jerusalem. A meetup on February 11, a part of OurCrowd official program, that will gather around 100 startup founders. Meetings of the participants on February 12 with Israeli companies interested in cooperation. And Ukraine’s stand at OurCrowd Global Investor Summit on February 13.</p><p>Ukraine Tech Days by Sigma Software is a good opportunity to unveil the potential of Ukraine, to get new connections, and partnerships.</p><p>Join us at the meetup, ping us if you want to meet with the participants at separate meetings, and/or visit Ukrainian stand at the OurCrowd Global Investor Summit in Jerusalem on February 13. More details on Ukrainian Tech Days in Israel are here:<a href="https://www.facebook.com/events/1831630260302617/"> https://www.facebook.com/events/1831630260302617/</a></p><p>Register to GROW WITH UKRAINE meetup at OurCrowd week: <a href="https://www.eventbrite.com/e/grow-with-ukraine-ukraine-meetup-ourcrowd-week-tickets-91752511231">https://www.eventbrite.com/e/grow-with-ukraine-ukraine-meetup-ourcrowd-week-tickets-91752511231</a></p><p><a href="https://www.facebook.com/events/1831630260302617/">The initiative is organized by Sigma Software with the help of </a><a href="https://www.facebook.com/UTEWtechTribe/?modal=admin_todo_tour">UTEW Tech Tribe</a> community, Ukrainian government institution<a href="https://www.facebook.com/exportpromotionoffice/"> Export Promotion Office</a>, backed by <a href="https://www.facebook.com/usaidcep.ua/">USAID</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c0f2e0cb120" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/3-winning-factors-for-successful-development-of-startups-6c0f2e0cb120">3 Winning Factors for Successful Development of Startups</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Becoming a Full Stack Engineer is a Stellar Idea?]]></title>
            <link>https://medium.com/sigma-software/why-becoming-a-full-stack-engineer-is-a-stellar-idea-72694bc2389d?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/72694bc2389d</guid>
            <category><![CDATA[full-stack]]></category>
            <category><![CDATA[cross-functional-teams]]></category>
            <category><![CDATA[full-stack-developer]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Mon, 04 Nov 2019 13:25:40 GMT</pubDate>
            <atom:updated>2019-11-04T13:27:43.344Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uzAD11eL_eN3RLLvenelzQ.jpeg" /></figure><p>Do you find cross-functional teams cool? Olga Paramonova, Senior Project Manager and Leader of one of the Competence Centers in Sigma Software, believes so. Over the last several years, her team took a course to switching to full stack. Why and was it of any help? What you, as an engineer, should get from taking a full stack side? Let`s find out.</p><p>Olga shares her experience of building a cross-functional team:</p><p>“For a significant period of time we stayed focused on establishing technology partnership and long-term relationships between our clients and engineering groups by building teams that can solve customer problems, think big, see products the same way as our customers do. Looking back at how my team started and where we are now, I can say that it was a huge step. We won a lot, it was not always a sunny day too, but we definitely saw a positive shift that both parties — customer and the team — could benefit from.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wgqAXZ4sDXOjcpR_IGePXQ.jpeg" /></figure><p>When we started, switching to full stack was not our plan from the first minute. However, as we learned our clients and understood their needs, it turned to be the only way.</p><p>It will be no surprise that clients think in terms of <strong>business needs</strong>, they rarely say, “Hey, can we have FE and BE team 50/50?” Most likely, they will ask you: “Hey, can we look into building this new platform for marketers, but in a more efficient way than the current ones?” And eventually, for sure, you will set up a team with frontend, backend, full stack engineers, you may even have a blast to afford designers. But take a look at this team a couple of years later. You will see that people and roles have significantly changed, and there are actually just full stack mates who are still with you.</p><p><strong>Why?</strong> If any of you stayed with your client from the very beginning to the launch day, you may know that to go live you need to fight the backlog. It can take as long as months of work, but sooner or later the important moment comes and, ta-da, the product you are working on is on the market! Now it is time to pause and see how the audience responds and there is not always a plan for even another month. During this period of uncertainty, your client needs people who can help further with anything, because they don’t know where exactly a new opportunity will light up. Thus, full stack guys now play a leading role and continue, while the rest of the team start looking for a new project. This is an unfortunate choice for all the parties, since the team you disband is already on the “performing” stage — people know each other and are all warmed up. They want to keep this experience, <strong>launch something cool again</strong>. But what can be done?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1cBmfKDyc-L3LWk1vplOsA.jpeg" /></figure><p>Actually, we found a way — risky, but it proved to be efficient. Two years ago, when we had to pass through this choice and decide who should leave, we took a challenge, and literally agreed with all fourteen teammates that no one should leave the team. Instead of reducing the capacity, we increased the number of businesses we could help. We asked our management to give us a second project, and because we had a wide tech profile, we found such project quickly. I think it worked well, we still have that core team with us as well as our first client!</p><p>However, we learned our lessons as well. A lot of our team members (over 80%) became full stack engineers, and most of our new mates, who are not there yet, have a roadmap to grow as FS. To reach the cross-functional playground, we worked hard to:</p><ul><li>Adjust our processes,</li><li>Close knowledge and experience gaps,</li><li>Revise leadership and team roles.</li></ul><p>Still, speaking about technology, I think that taking a full stack course played a key role for us. Today we are still on our path and we build our community in this way. Within two years, we turned a 14-person team into a 50+ Competence Center. It is always people’s work happening day by day to build the right mindset, explain corporate values, create individual roadmaps. We have ongoing internships for both engineers and managers to become a part of this community, start from small, but join the big. To keep our Senior group’s minds busy, we never wait to share leadership positions with them, and we are happy to have a flexibility here, we can just start any project.</p><p>To someone, the full stack expertise may seem too wide to be a professional in something particular, still, for us, its advantages are too heavy to ignore. Think about it — being a full stack engineer, you can:</p><ul><li><strong>Stay with your beloved team for a longer time</strong>;</li><li><strong>Create the entire thing that could alternatively need a team</strong>. And, surely you, having gained systematic thinking, can solve a wider range of problems. If we need to onboard an already existing solution, it is less stress and challenge, and less people to start;</li><li><strong>Speak solutions, not execute a part of the stories</strong>. Having this ability, you, as a professional, will be quickly noticed by the client and will have a great room for communication, improving your language and interpersonal skills (most of our engineers talk to clients directly, not via PM). When you have an open, direct communication with the client, you see how real the world is, no more illusions, and you question your decisions more often, becoming more mature and smarter;</li><li><strong>Choose between a wide range of opportunities to continue with something you love</strong>, since you have seen and tried a bunch of things and technologies. Moreover, that may be a good base for you to try yourself on leadership roles, such as TL or Architect, as you have to play with the whole system and understand how it gets wired up.</li></ul><p><strong>Are you ready to start?</strong></p><p>Make your choice, switch to full stack, join the light side of the force :-)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m64VkKPRBJ4oZ5o-xaY36A.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=72694bc2389d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/why-becoming-a-full-stack-engineer-is-a-stellar-idea-72694bc2389d">Why Becoming a Full Stack Engineer is a Stellar Idea?</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[PM’s Don’ts or How to Save Money, Reputation, and a Client]]></title>
            <link>https://medium.com/sigma-software/pms-don-ts-or-how-to-save-money-reputation-and-a-client-d1c8103c3e48?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/d1c8103c3e48</guid>
            <category><![CDATA[project-management]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Wed, 23 Oct 2019 10:36:40 GMT</pubDate>
            <atom:updated>2019-10-23T10:36:40.659Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rxy2Y2rlVlXUJ5ULfceoIg.jpeg" /></figure><p>Even seasoned Project Managers with dozens of successful IT projects behind them can slip. The problem is that, in the worst-case scenario, Project Managers’ mistakes entail disappointed clients, money loss, tarnished reputation, and client attrition. To avoid bitter mistakes, a good idea is to build on others’ experience. In Sigma Software, we made a practice of sharing experience between PMs at a yearly internal conference called <a href="https://sigma.software/about/media/pm-gathering-perfect-mix-valuable-knowledge-and-friendly-atmosphere">PM Gathering</a>. We used this year’s conference as an opportunity to meet the best of our experts in one place and ask them to share what PM’s mistakes they consider the most grave and how to avoid them.</p><h3>DON`T LOSE COMMUNICATION</h3><p>Olesya Khokhoulia, Deputy CEO, Account Management at Sigma Software warns against poor communication with a customer. “A very important indicator of your success is how often you communicate with the client. If you haven’t visited your client for 3–4 months and had no personal meetings during this time, then consider your chemistry dead. You have nothing to work with and are gradually turning into an email-sending machine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cIfklIrZ45Ko1yKzJ8cTjg.png" /></figure><p>How much communication is enough? If you meet your client every two months, then you are doing well. If the client tries to come to your office often, be the part of the team, hang out with them, then you rock!”</p><h3>DON`T NEGLECT EXAMINING THE CONTRACT</h3><p>How well do you know your project contract? Is your understanding of expectations bases on what your Account Manager told you? How well do you <strong>understand what lies behind the contract provisions</strong>? What are the areas of responsibility irrespective of the contractual conditions? What does the client actually expects from the project? The answers to these questions, Olesya says, define your successful execution of the project and whether you end it with income or loss. Don’t rely on hearsay and assumptions regarding your project. Learn a contract carefully and walk through the project progress, areas of responsibility, and expectations together with your client.</p><h3>DON`T HIDE BEHIND CONTRACT TERMS</h3><p>Anna Boiko, highly accomplished Account Manager at Sigma Software, is sure that using a contract to disengage from responsibility is another bad sign. A contract may grant you legal protection, but not a successful project completion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rPjJEg2FWabT3wLuHuXIKw.png" /></figure><p>If you get superficial requirements and estimates at the sales stage and draft a rough contract leaving room for re-estimation, you won’t meet client’s expectations. You may feel safe after staffing a team managed on the client’s side. However, a client may have certain expectations for a budget, and since you didn’t clear up the requirements and these hidden expectations, you cannot adjust their budget expectations. This may entail painful consequences over time. The client may run out of money and face the need to cut costs. They may be not ready to invest more and then will leave you disappointed. In this case, the project won’t make a hit, and you’ll suffer reputational damage.</p><h3>DON’T EXPECT YOUR CLIENT TO HANDLE RISKS</h3><p>There are no risks, says Olesya Khokhoulia. Feel puzzled? But think about it for a moment. What you are used to call a risk is actually a problem that needs to be solved. This problem may influence a customer or it may not. The main thing to remember is that in any case your client relies on you as a savvy supplier and expects that you will bring up all problems as early as possible, explain them clearly and accurately to enable their fastest elimination. You may mention a risk in your reports a hundred times, but if you do nothing about it, then you don’t fulfill your duties to the full extent. What you need to do is take necessary actions to eliminate the problem or prepare remedial measures to alleviate the consequences.</p><p>Let`s take a look at a real-life example. Have you ever took a flight? On a plane before takeoff, the crew always gives a safety training. Do you remember how many passengers pay attention to it? The answer is: quite few. But why? Because every passenger thinks: “I`ve chosen a good airline, I believe the pilot is a professional, so why should I worry?” Is there a chance that the plane crushes? Of course, there is. Do passengers perceive it as a problem that affects them personally? No, because they believe that all the problems will be solved by the people they relied on. The same thing happens in business.</p><h3>DON’T PUT STANDARD PROCEDURES AHEAD OF CLIENT’S NEEDS</h3><p>Anna Boiko is extremely good at finding the right solution and staying on the same page with a client. She is confident that a formal approach and know-it-all attitude prevents you to become a reliable partner for your customer.</p><p>“Fancy terms and long emails won`t make you easier to understand: your client doesn’t need your lectures. Choose a right communication style and avoid sounding like a snob or a bore. Listen to your client and really try to understand what they need. If your client seeks a creative solution, don’t push a ‘by-the-book’ approach accompanied by a preachment about practices. A tendency to follow the book may result from a fear to shoulder a responsibility. However, this is a highroad to lose creativity, stay with old practices, not learning new methods and not extending offer and expertise. Be keen to explore new ways, if you see that this is the best you can do.”</p><h3>ROUND-UP: WARNINGS TO NO-BEGINNER PMS</h3><p>After you’ve made your first mistakes in the project management trade, you may feel that you are on a roll. Keep in mind that you may still make mistakes, and they may be deeper and more adversely affecting the project outcome. So, we hope that the warnings in this article will help you detour the dangers and stay an efficient Project Manager leading your projects to success. Or maybe they will just confirm that you are on the right track.</p><p><em>Thanks to competent managers — Olesya Khokhoulia and Anna Boiko — for sharing their hard-earned practices.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d1c8103c3e48" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/pms-don-ts-or-how-to-save-money-reputation-and-a-client-d1c8103c3e48">PM’s Don’ts or How to Save Money, Reputation, and a Client</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cybersecurity for Developers: Cheat Sheet]]></title>
            <link>https://medium.com/sigma-software/cybersecurity-for-developers-cheat-sheet-c6a364298ad0?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6a364298ad0</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[secure-development]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Thu, 05 Sep 2019 12:58:50 GMT</pubDate>
            <atom:updated>2019-09-26T09:35:57.207Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Nickolay Mozgovoy is sure: the issue of cybersecurity must be a matter of concern not just for experts, but for ordinary developers as well. Not every project can afford a dedicated role of a security specialist, so the one to bare this burden, in one way or another, might be </em><strong><em>You</em></strong><em>. In the article below, Nickolay shares a cheat sheet he made up for such cases.</em></p><p>The created list is neither really short nor exhaustive, but at least it seems rather manageable and comprehensible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RkVmJw1dlyLqVBRQL7Gj0A.png" /></figure><p>I must confess that the material hereinbelow is ultimately inspired by a very nice course that I attended a few years ago and which I do recommend to everyone interested in the topic of InfoSec/Cybersecurity. So, be aware of a freely available <a href="https://www.coursera.org/learn/software-security">Software Security course offered by University of Maryland</a> on Coursera.</p><h3>Cybersecurity Cheat Sheet</h3><p>Software security is conceptually different and therefore not that intuitive compared to general functional requirements, of which we care foremost. Nevertheless, security is a crucial property of software, especially when it comes to the software-controlled machines that can affect the life and health of the customers or the systems that process personal data.</p><p>While usually the desired program behavior is treated as a primary goal, the <strong>main concern</strong> <strong>of security</strong>, in contrast, is what a program should not do, in other words, <strong>preventing the undesired behavior</strong>.</p><h3>Security Properties</h3><p>There are three primary properties of the software security, which we strive to keep from violation:</p><ul><li>Confidentiality</li><li>Means that sensitive information should not be leaked to unauthorized parties;</li><li>Integrity</li><li>Means that preserved information should not be damaged by unauthorized parties;</li><li>Availability</li><li>Means that a system should remain responsive to requests (according to the agreed conditions).</li></ul><p><em>While intuitively availability is associated with protection from </em><a href="https://en.wikipedia.org/wiki/Denial-of-service_attack"><em>DoS attacks</em></a><em>, it is not limited to them. Just consider the following problems of the same importance:</em></p><ul><li><em>Application upgrades;</em></li><li><em>IP address possession renewal;</em></li><li><em>Domain name possession renewal;</em></li><li><em>Encryption and authentication certificates rollover.</em></li></ul><p>Whenever such properties are violated, software is considered to have a <strong><em>vulnerability</em></strong>.</p><p><strong><em>Vulnerability</em></strong> is a security-relevant software defect that can be exploited to achieve undesired behavior.</p><p>If vulnerability is caused by a defect in <strong>design</strong> rather than code, it is called a <strong><em>flaw</em>.</strong></p><h3>Security Principles</h3><p>There is a set of common principles that you ought to follow when building secure software. In general, these principles can be split into the following several categories</p><h4>Principles to build software</h4><ul><li><strong>Prevention — eliminate defects entirely;<br></strong><em>For example, it is possible to eliminate entire sets of defects by using memory-safe languages like Java or C# instead of C/C++.</em></li><li><strong>Mitigation — reduce the harm from exploitation of an unknown defect;<br></strong><em>For instance, the harm of the hijacked database can be greatly reduced when the stored data is encrypted. Blocking accounts with suspicious behavior and requiring additional confirmation on potentially harmful actions is another example.</em></li><li><strong>Detection — identify and understand attack (i.e. Monitoring);<br></strong><em>It is necessary to get informed of potentially malicious actions as soon as possible. Comprehensive real-time telemetry makes a great help in such case.</em></li><li><strong>Recovery — undo the damage.<br></strong><em>“Quickly raised is not considered fallen,” therefore it is always necessary to have data backed up and extra resources ready to be available in case of urgency. In addition to that, it is highly important to have procedures of recovering hacked user’s accounts.</em></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v5npwFL4r8nB_FGe4ct6cg.png" /></figure><h4>Principles to design, implement, or maintain software</h4><p>The basic principles you ought to follow whenever you design, implement, or maintain software are as follows:</p><ul><li><strong>Favor simplicity (prevention);</strong></li></ul><ol><li>Favor simple design;<br><em>“</em><a href="https://en.wikipedia.org/wiki/KISS_principle"><em>Keep it simple, stupid</em></a><em>,” is a well-known meme that is utterly applicable in our context.<br>For instance, think twice when you choose between an old-fashioned monolith and overhyped microservices. Always consider available capabilities and expertise.</em></li><li>Use fail-safe defaults;<br><em>Never keep using default passwords, apply industry standard cryptography, favor whitelisting over blacklisting, etc.</em></li><li>Do not expect an expert user;<br><em>The more power you give in an unrestricted manner — the higher the probability of misusing it is. For example, installing applications from untrusted sources on Android (APK files) requires explicit activation of this feature in device settings, which is a fair foolproof solution.</em></li><li>Favor simple UI;<br><em>Again, complexity is the enemy of security. Complex UI increases the chance of a mistake from both developer’s and user’s sides.</em></li><li>Avoid allowing users to make security choices.<br><em>You cannot assume the users understand cryptography schemes or even rules of secure online behavior, therefore do not let them chose encryption algorithms &amp; keys or insecure passwords.</em></li></ol><ul><li><strong>Trust with reluctance (prevention/mitigation);</strong></li></ul><ol><li>Employ a small <a href="https://en.wikipedia.org/wiki/Trusted_computing_base">trusted computing base</a>;<br><em>The more components (both hardware &amp; software) your system consist of — the wider the attack surface is. In that sense, every new platform you are going to support and every service you integrate with should be considered as a source of risk and danger.</em></li><li>Avoid using your own cryptography;<br><em>Cryptography is such a non-trivial matter that trying to implement its aspects on your own is nothing but conceit, unless you’re not an expert. The right approach is to use open, industry-standard protocols and algorithms.</em></li><li>Least privilege possible for components and users;<br><em>Be careful when giving permissions to your application and users (even indirectly). Does your application really need access to a file system? Do read operations against a database need to be executed under a higher privileged account?</em></li><li><a href="https://cheatsheetseries.owasp.org/cheatsheets/Input_Validation_Cheat_Sheet.html">Input validation</a>;<br><em>Do not feed your system with malformed data. Clearly define the boundaries of valid data and enforce them by </em><a href="https://www.owasp.org/index.php/Data_Validation#Sanitize_with_Whitelist"><em>whitelisting</em></a><em>.</em></li><li>Promote privacy — limit access to personal data;<br><em>The formal definition of private/personal data may differ depending on the local law, while the common feature of private data is direct or indirect possibility to determine the actual person this data belongs to.</em></li><li>Compartmentalization — make use of processes, containers, and sandboxes to isolate components or even particular operations from each other.</li></ol><ul><li><strong>Defense in depth (all);</strong></li></ul><ol><li>Security by diversity (HTTPS + safe language + data encryption + VPN);<br><em>Combine all security mechanisms at your disposal. At the same time, omit using tools out of your competence and find an expert if necessary.</em></li><li>Use standard and open source solutions (avoid <a href="https://en.wikipedia.org/wiki/Security_through_obscurity">security by obscurity</a>);<br><em>Get rid of the idea that proprietary software or protocols (especially your own ones) provide any protection on the ground that nobody knows them. Only publicly available and studied tools and protocols can claim to be secure.</em></li></ol><ul><li><strong>Monitoring and traceability (prevention/recovery).</strong></li></ul><ol><li>Collect logs, telemetry;<br><em>Do not forget to create alarms on dangerous/suspicions events.</em></li><li>Make backups/snapshots.<br><em>Keep them on separate servers and have mechanisms of their recovery.</em></li></ol><p>For sure, it is possible to come up with more principles of that kind, while following the already defined ones should be sufficient in most of the cases.</p><h3>Secure Development Process</h3><p>It is quite common to process security issues reactively, deferring them until they become exploited and intolerable, which is a flawed approach.</p><p>Secure development process means that necessary activities and practices are introduced for each stage of development:</p><ul><li><strong>Requirements;</strong></li></ul><ol><li>Define security requirements (abuse cases);<br><em>Abuse case — is an opposite to a use case (or functional requirement). It should explicitly define what a system should not do.</em></li><li>Define required security properties of the system components (i.e. Confidentiality, Integrity, Availability);<br><em>Explicitly identify what parts of preserved data are confidential. Identify what roles are present in the system and what data they ought to manipulate. Arrange availability requirements.</em></li><li>Define security mechanisms to support these properties (<strong>Au</strong>thentication, <strong>Au</strong>thorization, <strong>Au</strong>dit);<br><em>Presence of these three </em><strong><em>Au*</em></strong><em> mechanisms is often regarded as a “Golden Standard of Security”. A system cannot be considered secure if it lacks one of them.</em></li><li>Define threat models.<br><em>This means a necessity to define a malicious actor and possible surfaces of attack (e.g. network traffic, local files, or even displayed data, etc.).</em></li></ol><ul><li><strong>Design;</strong></li></ul><ol><li>Threat-driven design (respond to previously defined threat models in system’s design);</li><li>Architectural risk analysis / assessment (identify potential flaws in design);</li><li>Apply security principles defined above (Prevention, Mitigation, Detection, Recovery).</li></ol><ul><li><strong>Implementation;</strong></li></ul><ol><li>Follow the best coding practice and guidelines;</li><li>Conduct mandatory code reviews;</li><li>Apply automation tools to enforce the highest code quality.</li></ol><p>i. <a href="https://en.wikipedia.org/wiki/Static_program_analysis">Static analysis</a>;</p><p>ii. <a href="https://en.wikipedia.org/wiki/Symbolic_execution">Symbolic execution</a>;</p><p>iii. <a href="https://en.wikipedia.org/wiki/Concolic_testing">Concolic execution</a>.</p><ul><li><strong>Testing;</strong></li></ul><ol><li><a href="https://en.wikipedia.org/wiki/Risk-based_testing">Risk-based</a> security tests;<br><em>These tests aim at the most critical parts of the system, identified by threat modeling.</em></li><li><a href="https://en.wikipedia.org/wiki/Penetration_test">Penetration testing</a>;<br><em>Security specialists play the role of the perpetrator in an attempt to overcome security mechanisms.</em></li><li><a href="https://en.wikipedia.org/wiki/Fuzzing">Fuzz testing</a>.<br><em>Feeding the system with random and malformed data may often expose undesired behaviors.</em></li></ol><h3>Common Design Flaws and Implementation Defects</h3><p>When following the secure development process, always keep in mind the fallacies widely committed by developers:</p><ul><li>Assume trust rather than explicitly give it;</li><li>Use an authentication mechanism that can be bypassed;</li><li>Authorize without sufficient context;</li><li>Confuse data and control instructions;<br><em>This is exactly the source of notorious </em><a href="https://en.wikipedia.org/wiki/SQL_injection"><em>SQL Injections</em></a><em>, </em><a href="https://en.wikipedia.org/wiki/Cross-site_scripting"><em>XSS attacks</em></a><em>, and </em><a href="https://en.wikipedia.org/wiki/Arbitrary_code_execution"><em>Remote code executions</em></a><em>.</em></li><li>Fail to validate data;</li><li>Fail to use cryptography correctly;</li><li>Fail to identify sensitive data;</li><li>Integrate external components without considering their attack surface;</li><li>Rigidly constrain future changes to objects and actors.</li></ul><h3>OWASP and Top 10 Vulnerabilities</h3><p><a href="https://www.owasp.org/">Open Web Application Security Project</a> is a not-for-profit organization focused on improving the security of software. And you can make a great use of its resources. This organization provides useful guidelines and “cheat sheets” for popular development platforms like <a href="https://www.owasp.org/index.php/Category:OWASP_.NET_Project">.NET</a> and <a href="https://www.owasp.org/index.php/Category:Java">Java</a>.</p><p>OWASP is primarily famous for its “Top Ten” lists describing the most common vulnerabilities in different domains:</p><p>· <a href="https://www.owasp.org/index.php/Category:OWASP_Top_Ten_Project">OWASP Web Top 10</a>;</p><p>· <a href="https://www.owasp.org/index.php/OWASP_Mobile_Top_10">OWASP Mobile Top 10</a>;</p><p>· <a href="https://www.owasp.org/index.php/OWASP_Internet_of_Things_Project">OWASP IoT Top 10</a>.</p><p>It is highly important to be well familiar with the common vulnerabilities of your domain: forewarned is forearmed.</p><h3>Security for Developers: the Gist</h3><p>Address security concerns first, do not dare to postpone them till the bitter end. Plan and carry out necessary activities through each phase of development to keep your customers safe and your reputation unquestioned.</p><p>I hope you’ll find this article helpful, but keep in mind that it is just a general cheat sheet. Becoming a competent specialist requires a lot of learning and practice.</p><p>You necessarily need to invest your time and resources to attend relevant courses or even achieve certification specific to the particular field of your interest.</p><p>Alternatively, if you are eager to pass on the burden of security concerns on the shoulders of professionals, you are heartily welcome to contact our highly skilled <a href="https://sigma.software/landing/it-security-services/">InfoSec team</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c6a364298ad0" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/cybersecurity-for-developers-cheat-sheet-c6a364298ad0">Cybersecurity for Developers: Cheat Sheet</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Top 5 Mistakes in an Interview]]></title>
            <link>https://medium.com/sigma-software/my-top-5-mistakes-in-an-interview-7cfccd597d47?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/7cfccd597d47</guid>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[recruiting]]></category>
            <category><![CDATA[interview-mistakes]]></category>
            <category><![CDATA[interviewing]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Mon, 12 Aug 2019 14:29:13 GMT</pubDate>
            <atom:updated>2020-06-18T15:26:38.521Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong><em>My Top 5 Mistakes as an Interviewer</em></strong></h3><p>Sergey Lysak, Project Manager at Sigma Software, has conducted dozens of interviews trying to hire best professionals to his team. Looking back at all those conversations, he now sees how much he learned over the years and shares his hard-earned lessons with you.</p><p>Sergey Lysak:</p><p>“Some smart books say that admitting a mistake is the first step to putting it right, others say it’s impossible to learn a lesson without making mistakes. I guess I am an expert in making mistakes and then relieving the consequences. I am not so sure about the consequences, but anyway, here is my list of top 5 mistakes I have made during interviews.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uppeSazWWu_vuJZKxGR02g.jpeg" /></figure><p><strong><em>Mistake N1 — We choose, but are not chosen</em></strong></p><p>In the process of getting ready to an interview, I used to look at it as a process where my task is to find a person best suited to the vacancy. It was the only aim I kept in mind during both the preparation and the interview itself.</p><p>I would come to an interview so self-important and full of myself, feeling ‘They must take me the way I am’. When the recruiter who set up the interview introduced me as an expert to conduct the technical part, I would look like a picture of ‘Got it? Did you hear? I am an Expert, and you are NOT, haha!’</p><p>And then I would be baffled when the recruiters shared a feedback that the candidate didn’t accept the offer for some odd reasons or gave a negative feedback about the interview.</p><p>Of course, I do not think that all offers were rejected for this reason, but some of them surely were. Leaving the interview room, the candidates asked themselves, ‘Maybe they are all like this here?’ or ‘Damn, will I have to work with this weirdo?’ These questions definitely cannot be regarded as a good result of an interview.</p><p>When this smug feeling died away, I realized that any interview is a two-way process. As someone who has to conduct an interview, I must remember that just like the company is choosing an employee, the candidate is choosing a company to work for. And neither of the parties has a right to look down on the interests of the other one.</p><p>The candidate’s aim is to sell oneself. The interviewers’ aim is to sell the company/project/team.</p><p><strong><em>Mistake N2 — Dominance in expertise</em></strong></p><p>I won’t hide it, when I was first asked to conduct an interview, it was so exciting. It seemed to me like a certain recognition of my skills by my boss and by the company as a whole. I felt my own importance. Of course, I wanted to demonstrate my best.</p><p>I was preparing for the interview not only through studying the vacancy’s requirements and the project’s nuances, but also through reading a ton of articles about the areas, of which I had no clear or structured knowledge myself. The more interviews, the more knowledge I acquired. Of course I wanted to apply this knowledge — preferably right there and then, at the interview.</p><p>What did it come to? To me asking people unnecessary questions or digging too deep in details. Not because it was important, but just because I had a chance, a wish, and the knowledge. After all this I drew conclusions, now I think mostly wrong ones.</p><p>Looking back, it seems sort of amusing and illogical — I would seek and find knowledge, take the interview where I could show it off, and then feel indignant that the candidate didn’t have the same knowledge.</p><p><strong><em>Mistake N3 — A relay race</em></strong></p><p>Before each interview I would look at the check-list that I had to use for drawing results. Quite often, it contained 15–20 lengthy points. I always had a feeling that I had to give a say on each of them, and back it up with arguments. It turned the interviews into relay races, which looked more like a ‘Mindfight’ quiz show — the candidate only missed the red answer button. It was superimportant for me to ask about each point on the check-list. So, all other parts of the interview were more and more non-existent, I wanted to ask more and more, and combined with mistake N2 it transformed into an endless digging session.</p><p>Now I prefer to have a list of priorities, where I try to figure out, which conclusions I have to make and which are optional. It makes no sense to set a ‘cover all’ goal — 1.5–2 hours will be never be enough.</p><p><strong><em>Mistake N4 — Demonstrating negative emotional response</em></strong></p><p>I would wrinkle my nose, roll my eyes, and shake my head while a candidate was talking. Sometimes I even noted down minus signs on a paper without any explanation. It took me way more than one interview to realize that these negative emotional reactions were the reason for the interview fading out on a third to fifth question, even if it had started in a positive and high-spirited way. Such reaction gives the candidate an idea that they have already failed, so I cannot count on any more thoughtful answers.</p><p>Therefore, I say no to negative emotional responses and yes to positive emotional responses.</p><p>By positive responses I mean nodding to a correct answer, a subtle joy to a found solution, using ‘positive’ and intensifying adjectives — anyone has something to praise for. The strongest ones are to admit that this idea has never come to my mind, or that I have never heard about that approach.</p><p>Now I believe that the best reaction to a wrong answer is a mere statement, if necessary. Ideally, paired up with the right answer or at least giving a direction for the candidate where to dig.</p><p><strong><em>Mistake N5 — Comparing opinions</em></strong></p><p>Even when I asked an open question, like ‘What would you do if the release was tomorrow and you had three critical points that still need fixing?’ I had my own opinion what to do. This opinion was usually shaped up to the smallest steps. So, I used to forget that it was only an opinion, and it was only one of a million ways to approach the question.</p><p>One big difficulty for me was to listen and follow the candidate’s flow without comparing their opinion with mine every other second, without finding differences and trying to convince the candidate that my opinion is right while theirs isn’t.</p><p>The point of these questions is to start a dialogue with an aim to understand the candidate’s way of thinking, to see what additional questions they ask and what things they notice. There is no point in comparing opinions ‘I think you have to report the issue and plan the fixes’ and the candidate’s ‘I think you have to keep fixing without telling anyone’, when the situation is completely hypothetical.</p><p>I did not write this with an aim to make a confession; however, maybe for this too. My main point was to help others use my mistakes and lessons learnt, or at least to admit they screw up just like I do.”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7cfccd597d47" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/my-top-5-mistakes-in-an-interview-7cfccd597d47">My Top 5 Mistakes in an Interview</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Benefits of Using UI First: Real-Life Experience]]></title>
            <link>https://medium.com/sigma-software/benefits-of-using-ui-first-real-life-experience-ed81814b028e?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ed81814b028e</guid>
            <category><![CDATA[ui-design]]></category>
            <category><![CDATA[storybook]]></category>
            <category><![CDATA[ui-first]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[ui-design-for-mobile-apps]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Tue, 30 Jul 2019 09:59:25 GMT</pubDate>
            <atom:updated>2019-07-30T09:59:25.800Z</atom:updated>
            <content:encoded><![CDATA[<p>This is a story told by Sergey Bakaenko about him and his team creating UI with UI First in mind. The approaches they used can be applied to any platform in general and will be of interest to most developers who create UI in their applications. Besides, at the end of the article, you can find attached some examples of how the described approaches were implemented for <a href="https://forma.fit/">an Android project.</a></p><p>Sergey Bakaenko tells:</p><p>Recently we have finished development of an MVP Android application. We managed to stay within the estimates and the scope drawn up at the start of the project. During the development, the number of UI bugs was insignificant and the review of our application by the design team was highly positive.</p><p>Based on the results of the work done, I would like to share some things that, as we see it, have allowed us to achieve high quality of the final product.</p><h3>Developing an Android App with UI First Approach</h3><p>At the beginning of the development, we decided to use the UI First approach. The main idea of this approach is that when we start working on a new feature, we first create a UI for it, test the implementation on the test data, collect feedback from the team, designers and, ideally, the customer, and only after this we get down to the implementation of the business logic for this feature.</p><p>This approach allows achieving the following results:</p><p>1. To detect various states of our screen, which are sometimes left out in the <a href="https://sigma.software/services/ui-ux-design#utm_source=medium&amp;utm_medium=post&amp;utm_campaign=UI-First">UI/UX development</a>. This way, fewer bugs related to invalid states or unprocessed cases in UI go to test engineers, because the developer is able to detect them at the stages of implementation and dev testing.</p><p>2. To provide designers with a capability to assess their design on a real device, so they can understand how it looks and ‘feels’ and make some changes if needed.<br> We all know that static mockups are one thing, while real design on a device gives a completely different feeling.<br> This way, the feedback from our design team is not put off to the late stages of the project. This helps to receive and apply the designers’ feedback in small increments. <br> These small chunks are easier to review for designers, while developers spend less time on fixing comments after the review.</p><p>3. One of the most important things is that this approach allows the customer to assess the UI of a new functionality before any considerable amount of time is spent on it. Any changes at this stage will be a lot less costly than when the business logic and the tests for this screen are finished. Since if customer decides to change the UI significantly, all the business logic also needs to be modified, which requires a lot of effort and time. To implement this approach effectively, we have created a Storybook for our application.</p><h3>Storybook for our Mobile App</h3><p>To look through various states of the UI without rebuilding the project, we were in need of an appropriate tool.</p><p>This is where the Storybook came in handy. The Storybook is a list of all screens (modal windows, standalone views, if necessary).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*dNXpCArC9eewMP0Q2aA1fA.png" /><figcaption>A sample view of the Storybook in our application</figcaption></figure><p>After selecting one element, we can select the state of this screen that we are interested in and then see the screen itself in the state selected earlier.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*Tx7aq-2VdMsH3oAWTAN6pw.png" /><figcaption>Selecting a state for the Onboarding screen</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*l6hDqcvC62NzYP_Epyxf8A.png" /><figcaption>The Onboarding screen with test data</figcaption></figure><p>This approach enables us to see various states of each screen without restarting the application. It also allows us to implement features faster, as opening a certain screen requires us to put the application into a certain state. This might be rather time-consuming as you need to do it after each change and build of the project.</p><p>Therefore, with the help of the Storybook, any screen can be reached in two clicks, without performing any other related actions. <br> <br> The idea of creating a Storybook is not new. An example of its implementation for web development can be found at this <a href="https://storybooks-official.netlify.com/?path=%2Fstory%2Fui-panel--default">link</a>. We adapted it for mobile technologies to ensure the viability of this solution and to understand what advantage it can give us.</p><p>The selected development method requires certain approaches to the architecture. In particular, we used Data Driven UI, which enables giving real or test data to the screen. At the same time, our UI is absolutely unaware of which data it is showing. This approach ensures the correspondence of the design that we create within the Storybook with test data and design with real data.</p><p><strong>Architecture of the App</strong></p><p>The aim of this article was to describe the overall approaches, without touching the specifics of the Android frameworks that we used.</p><p>Here is a <a href="https://github.com/sergeybakaenko/UI-First-Sample">link to a github project</a> that is essentially a replica of all frameworks and approaches used at our real project. There you can also find the detailed architecture documentation. The provided example project includes only one screen showing a list of all repositories of github API.</p><p>More articles about software design and development <a href="https://sigma.software/about/media#utm_source=medium&amp;utm_medium=post&amp;utm_campaign=UI-First">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ed81814b028e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/benefits-of-using-ui-first-real-life-experience-ed81814b028e">Benefits of Using UI First: Real-Life Experience</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Smarter Way to Source Ideas in Corporate Environment]]></title>
            <link>https://medium.com/sigma-software/a-smarter-way-to-source-ideas-in-corporate-environment-8a7fa8b75417?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/8a7fa8b75417</guid>
            <category><![CDATA[research-and-development]]></category>
            <category><![CDATA[ideas]]></category>
            <category><![CDATA[image-recognition]]></category>
            <category><![CDATA[hackathons]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Fri, 28 Dec 2018 13:46:49 GMT</pubDate>
            <atom:updated>2018-12-28T13:46:49.828Z</atom:updated>
            <content:encoded><![CDATA[<p><em>In order to get non-trivial results, you need to think out-of-the-box. Volvo Construction Equipment decided to organize VOLVO x HACK SPRINT Hackathon to source ideas and potential solutions to an existing business problem as opposed to a traditional approach in developing new solutions. A team from Ukraine representing </em><a href="https://sigma.software/#utm_source=Medium&amp;utm_medium=article&amp;utm_campaign=VCE-Hakathon"><em>Sigma Software</em></a><em>, part of Sigma Group, a consulting group established by Swedish entrepreneur and philanthropist Dan Olofsson, won the Hackathon demonstrating that consulting companies are open to fresh ideas, outside-the-box approaches and expert knowledge alongside traditional consulting services.</em></p><p>On November 23–26, Volvo Construction Equipment ran Volvo x Hack Sprint dedicated to image intelligence technologies. The Hackathon attracted 15 pre-selected international teams that gathered in Lindholmen to create innovative solutions to an actual business problem suggested by Volvo CE — generate meaningful data on machines and help take proactive actions regarding their availability.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JtuhjW3irZeA_HSxemd10A.jpeg" /></figure><p>Volvo CE ambition was to use Hackathon as a new and innovative approach to sourcing ideas and finding the best solutions. The advantages were obvious: involving people with different backgrounds and different skillsets to get a broader view of the problem looked like a perfect prerequisite to getting an out-of-the-box vision. Another advantage of this approach is that instead of just getting ideas shaped into specification, Hackathon would result in actual PoC solution one could use as a basis for further development.</p><p>“I am impressed with how people approached the problem,” expresses Mya Istanbouly, Business Analytics Services Leader at Volvo Construction Equipment who was among the ones inspiring the event and organizing it. “Everyone I have spoken to looked at the problem from a different perspective and all created solutions were extremely intelligent. This made the winner selection a really tough task.”</p><p><a href="https://sigma.software/about/media/sigma-software-team-wins-volvo-x-hack-sprint#utm_source=Medium&amp;utm_medium=article&amp;utm_campaign=VCE-Hakathon">First prize went to the Sigma team from Ukraine</a>, which proposed the solution at the intersection between image intelligence and machine learning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NCPuRscq1r3OpRm4BV83dQ.jpeg" /></figure><p>“Participating in the Hackaton was an interesting experience for us, as it goes beyond traditional consulting activities where you already have the idea of potential solution defined by the stakeholders. Our domain expertise indeed contributed to deeper understanding of the business problem,” reflects Tetiana Osetrova, Sigma Software Hackathon Team Leader. “However, this was not the only success factor. All our team members had cross-expertise background from projects delivered to other industries and business domains. Thus, we managed to combine automotive domain knowledge with insights from other industries. Moreover, we have a lot of R&amp;D initiatives at Sigma following most recent technology trends. Experience from those also helped a lot.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FxkQZzSROd9MuvmDTAWHcw.jpeg" /></figure><p>Historically, most companies would try to define a business problem, brainstorm internally to elaborate a potential solution, and then turn to a consulting company to get this solution formalized and developed. However, recent industry developments show that most consulting companies shift their focus to becoming technology partners, coming up with new ideas and solutions. When you consider relatively big IT consultancies, most of them would have substantially increased the volume of R&amp;D activities within the last three years.</p><p>“We work in a highly competitive market and this competition pushes companies to differentiate,” acknowledges Artem Petrenko, Managing Director of Sigma Sweden Software AB. “You can no longer differentiate by just offering software development. Well-established processes, quality practices and technical excellence are the major differentiators today. When you think long-term, the focus shifts towards being able to add value to your customer’s business. This is why we now actively invest in <a href="https://sigma.software/about/media/diving-deep-rd-expert-interview">R&amp;D laboratories</a>, in which our experts play with AI, AR/VR, IoT, Big Data, and so forth, as well as into creation of background intellectual property to be applied to our customers’ projects.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d88O0jphltbLMtWIDrqY8A.jpeg" /></figure><p>The next time you contemplate what and how to outsource, remember those trends. When you want more – ask for more.</p><p><strong>About Volvo Construction Equipment:</strong></p><p>Volvo Construction Equipment is among the world’s leading manufacturers of articulated haulers and wheel loaders, and one of the world’s foremost manufacturers of excavation equipment, road development machines and compact construction equipment. As a total-solution provider, Volvo Construction Equipment also offers servicing, financing, used equipment, rental and other related services.</p><p><strong>About Sigma Software:</strong></p><p>Sigma Software is a part of Sigma — a group of leading consulting companies with the objective to make our customers more competitive. Sigma Software provides top-quality software engineering services for global corporations, software product houses and start-ups in Scandinavia, Europe, and the US with special focus on technical excellence and innovation.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8a7fa8b75417" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/a-smarter-way-to-source-ideas-in-corporate-environment-8a7fa8b75417">A Smarter Way to Source Ideas in Corporate Environment</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automatic question answering]]></title>
            <link>https://medium.com/data-science/automatic-question-answering-ac7593432842?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ac7593432842</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[information-retrieval]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[neural-networks]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Mon, 12 Nov 2018 15:00:54 GMT</pubDate>
            <atom:updated>2018-11-13T11:08:39.221Z</atom:updated>
            <content:encoded><![CDATA[<h3>Automatic Question Answering</h3><p>Querying Information from structured and unstructured data has become very important. There is a lot of textual data, FAQ, newspapers, articles, documentation, user cases, customer service requests, etc. Quite hard to keep in mind all that information. Who won the last football Euro Cup? What is Bitcoin? When is the birthday of a famous singer? To find some information, we need to search a document, spend a couple of minutes reading before you find the answer. Nazar Grycshuk, Data Scientist at Sigma Software and the author of this article, believes that automatic QnA system could save the most valuable resource in our lives — time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qUFgGhSERoWAa08MV6AVCQ.jpeg" /></figure><p>First, let’s discuss what types of question are there, and what answers to such questions users can expect.</p><p>All questions can be divided into several types:</p><ul><li>General questions, with Yes/No answers</li><li>Wh-Questions, start with: who, what, where, when, why, how, how many</li><li>Choice Questions, where you have some options inside the question</li><li>Factoid questions, where the complete answer can be found inside a text. The answer to such questions consist of one or several words that go one after another</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ly0p8VfcqVSRMgbnfNT7UQ.jpeg" /></figure><p>In this article, we will focus on the factoid question answering approach that can process a wide range of questions and covers one of the most difficult type of them.</p><p>The approach that will be discussed in this article is based on an algorithm designed specifically for factoid question answering. As to other question types (general, Wh-questions, choice questions), the algorithm could be applied after some adjustments. Also, some ideas could be reused to create your own approach.</p><h3>Main Factoid Question Answering Approach</h3><p>Typically, a question answering system consists of several logical blocks:</p><ul><li>Data source,</li><li><strong>Information retrieval (IR) system</strong>,</li><li><strong>Machine reading comprehension model (MR model),</strong></li><li>One or more additional blocks, such as modules for text pre-processing, answers post processing, checking, and stabilization.</li></ul><p>Let’s see how these blocks interact with each other:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/682/1*-2ipuxL7bjKfnEkvJX6dRQ.png" /></figure><p>As you can see in the picture, when the question is transferred to the system, the first step to be done is the search of the texts that can contain the answer to the question. The sources of such texts can include various web pages, documents stored on a local disc, Wikipedia, etc.</p><p>The next step is to select the most relevant texts from all the information that was found. For this purpose, the following approaches could be used: TF-IDF similarity, Jaccard index, Word embeddings, Deep Learning text similarity algorithms. All of these approaches have pros and cons, different time/memory needs for pre-processing and getting similarity index. There is no silver bullet or best approach for all cases. Which approach is the best for a concrete task depends on several factors like domain area, dataset, amount of data. The problem is complex and has many facets, every month a couple of articles about IR/Text similarity task are published.</p><h3>Information Retrieval System</h3><p>According to Wikipedia, “an open domain question answering system aims at returning an answer in response to the user’s question. The returned answer is in the form of short texts rather than a list of relevant documents.”</p><p>If you choose to answer open-domain questions, the source of information can be: your own documents, Common Crawl, Wikipedia, some ontologies like DBpedia or Google knowledge graph, also you can try Google custom search engine.</p><p>If you decide to use Wikipedia, there are 2 approaches. The first one is searching in downloaded Wikipedia pages. The second one is extracting entities (e.g. person’s name, location, dates) from a question and then extracting all wiki pages with those entities. However, there is a possibility that a Wiki page with the entity does not cover your question, and you will not find the answer in the text.</p><p>In our project Google custom search engine was chosen, because it returns the most relevant information not only from Wiki, but also from other indexed pages. Search could be adjusted by adding special parameters, location, topics, website types, etc. Also, Google custom search engine returns a list of URLs with a short summary, that could be used as an answer passage for the next steps.</p><p>Here is an example of an automatically found answer to the same question by using different approaches.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/975/1*qugdspNJAq1RNkGgajLxTQ.png" /></figure><p>At the previous step, we end up with a list of relevant documents. The next step is to split documents into passages and find the most relevant ones.</p><p>The following algorithms were used:</p><ul><li>Ranking of passages with tf-idf/Jaccard/cosines distance based on word embeddings</li><li>Deep Learning (DL) approach like Bi-LSTM model with avg pooling and attention (https://arxiv.org/pdf/1511.04108.pdf)</li></ul><p>Depending on data, 0.63–0.85 F1 score could be archived. As a result of this step, we will get a list of passages sorted by a similarity index. Let’s take 5–10 most relevant to the next step.</p><h3>Machine Reading</h3><p>The last step is the most interesting one. We have a question and a passage. The idea is to extract the answer from the passage. There is a bunch of approaches that can be applied here. For example, you can build an abstract representation for the question and passage, get a question type (so you will know the answer type), and get an answer. But in practice, this approach is quite complicated even for close domain question with restricted questions types and entities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tjq_iRwvYZf4uox1V1-ZwA.png" /></figure><p>Most of the modern approaches are based on Deep Learning. To use these approaches, we need to have datasets. There is a list of datasets, such as SQuAD, CNN\Dayli Mail, MSMarco, TriviaQA, WikiReading, to train neural network (NN).</p><p>We’ve trained our model on <a href="https://arxiv.org/abs/1606.05250">SQuAD dataset v1</a> and <a href="https://arxiv.org/abs/1806.03822">v2</a>. V1 contains 100,000+ question-answer pairs with content from 500+ documents related to different topics. In v2, there were added 50K questions without answers. The models trained and tested on this dataset can be compared via a <a href="https://rajpurkar.github.io/SQuAD-explorer/">leaderboard</a>. The structure of the dataset looks as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X5xTuj8lHyqrnwyHnUj5hA.png" /></figure><p>There is another type of datasets, where the answer to the question is not in the context. An example of such dataset type is MS Marco. There is 1M+ Bing user query, 160К+ Answers.</p><p>Here is a list of good NN architectures for Question Answering that can be helpful in creating a question answering system.</p><ol><li>AoA <a href="https://arxiv.org/abs/1607.04423">https://arxiv.org/abs/1607.04423</a></li><li>r-NET <a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2017/05/r-net.pdf">https://www.microsoft.com/en-us/research/wp-content/uploads/2017/05/r-net.pdf</a></li><li>ReasoNet <a href="https://arxiv.org/abs/1609.05284">https://arxiv.org/abs/1609.05284</a></li><li>FusionNet <a href="https://arxiv.org/abs/1711.07341">https://arxiv.org/abs/1711.07341</a></li><li>HyperQA <a href="https://github.com/tjurczyk/HyperQA/blob/master/LICENSE">https://github.com/tjurczyk/HyperQA/blob/master/LICENSE</a></li><li>BiDAF <a href="https://allenai.github.io/bi-att-flow/">https://allenai.github.io/bi-att-flow/</a></li><li>RDF <a href="https://www.semanticscholar.org/paper/Natural-language-question-answering-over-RDF%3A-a-Zou-Huang/cb4d277a51da6894fe5143013978567ef5f805c8">https://www.semanticscholar.org/paper/Natural-language-question-answering-over-RDF%3A-a-Zou-Huang/cb4d277a51da6894fe5143013978567ef5f805c8</a></li><li>Reinforcement Mnemonic Reading <a href="https://arxiv.org/pdf/1705.02798.pdf">https://arxiv.org/pdf/1705.02798.pdf</a></li></ol><p>For our task, R-Net model was used. The choice was made in favor of this model because of its good score on state-of-the-art leader boards and a well-described architecture. The R-Net model is based on Bidirectional Recurrent Network where question and passage are processed separately. After that, a gated attention-based recurrent network matches a question and a passage to get a question-aware representation for the passage. Then, the self-matched attention aggregates evidences for the whole passage. The output layer predicts the start and end positions of the answer inside the passage.</p><p>Image from <a href="https://www.microsoft.com/en-us/research/wp-content/uploads/2017/05/r-net.pdf">www.microsoft.com/en-us/research/wp-content/uploads/2017/05/r-net.pdf</a>.</p><h3>Answers to Random Questions</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/976/1*E7fAkKuVUVYukNqN8TZhVw.png" /></figure><h3>Our Test on Factoid Question Dataset: 62%(EM) Exact Match</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/944/1*cwbtLY3gvTIk9SiQFuHpxA.png" /></figure><h3>Summary:</h3><p>We end up with quite good machine reading model in combination with Wiki-based and Google custom search Information retrieval system. The system quality can be improved by combining different information sources, post processing with combining and scoring results, train new architectures and build ensemble model, increase dataset for training MR model with some specific cases.</p><p>This approach can be applied for real business cases as a part of Question answering/user support system. If you are building a question-answering system and use NLP engine, like Rasa NLU, Dialogflow, Luis, this NLP engine can answer predefined questions. However, if there is no predefined intent, you can call this automatic QnA system to search in documents and return the answer. Continuous learning can be organized by getting feedback from users. For now, the quality of this model can’t outperform people, but it is really close to it (70–80%) depending on a dataset. Every month the quality of such models increases. It means that in the very near future such models may answer questions with the same accuracy as humans, but much faster.</p><p>I hope this article will help you in creating your question answering system without spending time on trying non-optimal approaches or tools.</p><p><strong>Sigma Software AI expertise</strong></p><p>Sigma Software gathered AI experts under its roof and supports them in investigating specifics of machine learning, computer vision, natural language processing, robotics, and other aspects of AI. The experience our experts get in such a way they apply in real-world projects and products ranging from <a href="https://sigma.software/case-studies/intelligent-chatbot">chatbots</a> to automated <a href="https://sigma.software/case-studies/procurement-planning-solution">procurement planning solutions</a>. Learn more about Sigma Software <a href="https://sigma.software/landing/artificial-intelligence/">AI expertise</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac7593432842" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/automatic-question-answering-ac7593432842">Automatic question answering</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GP-GPU Computing via C#]]></title>
            <link>https://medium.com/sigma-software/gp-gpu-computing-via-c-7ff49312df43?source=rss-da4ff3d1388c------2</link>
            <guid isPermaLink="false">https://medium.com/p/7ff49312df43</guid>
            <category><![CDATA[csharp]]></category>
            <category><![CDATA[gpgpu]]></category>
            <category><![CDATA[gpu]]></category>
            <category><![CDATA[cuda]]></category>
            <category><![CDATA[dotnet]]></category>
            <dc:creator><![CDATA[Yana Arbuzova]]></dc:creator>
            <pubDate>Wed, 18 Jul 2018 14:11:29 GMT</pubDate>
            <atom:updated>2018-07-18T14:11:29.157Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Almost every year we increase the number of CPU cores in our devices to increase overall performance and user experience. Having an eight-core phone is not a big deal nowadays. Although there exists another kind of programmable unit that is usually ignored by most of the programmers. It has a multitude of computing cores, hundreds of them, and it is a GPU or graphics processing unit responsible for drawing the user interface and handling 3D experience in games. From the very beginning of its existence, it was a highly specialized device designed just for transforming and rendering the given data, and there was only one-way flow of data: from CPU to GPU.</em></p><p><em>However, since the arriving of Nvidia CUDA (Compute Unified Device Architecture) in 2007 and OpenCL (Open Computing Language) in 2009, the graphics processing units became accessible for general-purpose, bidirectional computations (called General-Purpose GPU Programming or simply GPGPU).</em></p><p><em>Nick Mozgovoy shares his perspective as a .NET developer regarding a great opportunity to have access to a huge computational power of hundreds GPU cores, so I tried to figure out what is the current state of art in my domain.</em></p><p>First of all, what CUDA and OpenCL are and what are the differences between them?</p><p>In general, they are APIs that allow a programmer to perform a specific set of computations on GPU (or even exotic devices like FPGA). It means that instead of rendering the result on display, the GPU will somehow return it to the API caller.</p><p>There are considerable differences between the two technologies.</p><p>Firstly, CUDA is a proprietary framework developed and supported only by Nvidia, while OpenCL is an open standard rather than a complete solution or concrete implementation. Therefore, CUDA is available only on Nvidia devices, while any manufacturer may support OpenCL (by the way, Nvidia chips support it as well).</p><p>Secondly, CUDA is a GPU-specific technology (at least now), while OpenCL interface may be implemented by various devices (CPU, GPU, FPGA, ALU, etc.).</p><p>These differences have obvious consequences:</p><ul><li>CUDA is a little bit more performant than OpenCL on Nvidia chips;</li><li>You certainly can rely on consistence between CUDA documentation and implementation having a single manufacturer (Nvidia), which is not the case with OpenCL;</li><li>OpenCL is the only way to go if you have to support hardware other that Nvidia chips.</li></ul><h3>HOW IT WORKS</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/711/0*kVQMEuKXWn9fJPDG.png" /></figure><p>Image: CUDA processing flow</p><p>Let us describe how GPGPU works with the scheme represented in Image 2:</p><ol><li>Form the data to be processed in RAM</li><li>Copy processing data into video RAM</li><li>Instruct GPU to process the data</li><li>Execute in parallel on each core</li><li>Copy the result back to RAM</li></ol><p>It should be noted that this kind of general-purpose GPU computations is reasonably restricted:</p><ul><li>they cannot perform any IO;</li><li>they cannot directly reference data in computer memory.</li></ul><p>Even though it seems simple on general scheme, the computation model and API are not that intuitive, especially considering the fact that the native API itself is available only in C and C++.</p><p>Methinks it is the main reason why the GPGPU programming is not really widespread yet.</p><h3>GPGPU ON .NET PLATFORM</h3><p>There is no native support of GPU programming on .NET platform yet, so we will have to rely on third party solutions. Moreover, there are not that many options to choose from, so let us briefly review available alternatives among actively developing projects. Interestingly, most of them focus on Nvidia CUDA rather than OpenCL.</p><p>Read the <a href="https://sigma.software/about/media/gp-gpu-computing-c#utm_source=medium&amp;utm_medium=report&amp;utm_campaign=article_GpGPUviaC">full article</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7ff49312df43" width="1" height="1" alt=""><hr><p><a href="https://medium.com/sigma-software/gp-gpu-computing-via-c-7ff49312df43">GP-GPU Computing via C#</a> was originally published in <a href="https://medium.com/sigma-software">Sigma Software</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>