<?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 Tech@ProSiebenSat.1 on Medium]]></title>
        <description><![CDATA[Stories by Tech@ProSiebenSat.1 on Medium]]></description>
        <link>https://medium.com/@prosiebensat-1?source=rss-b7364a9a56b1------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*PfjiDGFK9P_4KFFmqZzIww.png</url>
            <title>Stories by Tech@ProSiebenSat.1 on Medium</title>
            <link>https://medium.com/@prosiebensat-1?source=rss-b7364a9a56b1------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 21:44:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@prosiebensat-1/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[From Keywords to Vectors: How Glomex Built a Contextual Video Matching Engine at Scale]]></title>
            <link>https://medium.com/tech-p7s1/from-keywords-to-vectors-how-glomex-built-a-contextual-video-matching-engine-at-scale-a9cfa54ee85e?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9cfa54ee85e</guid>
            <category><![CDATA[elasticsearch]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[information-retrieval]]></category>
            <category><![CDATA[semantic-search]]></category>
            <category><![CDATA[system-architecture]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Wed, 17 Dec 2025 15:22:26 GMT</pubDate>
            <atom:updated>2025-12-18T14:23:50.314Z</atom:updated>
            <content:encoded><![CDATA[<p><em>By </em><a href="https://www.linkedin.com/in/shamilyagiyayev/"><em>Shamil</em></a><em> </em><a href="https://medium.com/@shamil.yagiyayev"><em>Yagiyayev</em></a><em> and </em><a href="https://www.linkedin.com/in/popadko/"><em>Oleh </em></a><a href="https://popadko.medium.com/"><em>Popadko</em></a></p><p>The engineering journey from rigid rules to hybrid semantic search delivering 2 billion views.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x-qZkkUhbPkfKSfELaF8mw.png" /><figcaption>Teaser is generated with AI by the Authors</figcaption></figure><h3><strong>Introduction</strong></h3><p>At Glomex, a <strong>ProSiebenSat.1 Media SE </strong>subsidiary, we operate a B2B video marketplace that connects content owners with publishers. Our core value proposition is a seamless “video-as-a-service” model: content owners upload premium video assets, and publishers (like news websites) integrate them to generate ad revenue via our Player solution.</p><p>For our publisher partners, the goal is to enrich as many articles as possible with a video that maximizes user engagement and revenue. Ideally, this is a <strong>Perfect Match — </strong>a specific video directly relevant to the article’s story (e.g., a match highlight for a sports report). These high-confidence matches are typically placed prominently at the top of an article. However, if a direct hit isn’t available, the system may fall back to a <strong>Related Match</strong> (e.g., section “You may also be interested in…”) placed further down the page. It is critical for our solution to accurately distinguish between these two states to ensure the video aligns with the article’s narrative and creates the right experience for the reader.</p><p>We designed an automated solution to support publisher editorial teams not just at the moment of publication, but throughout the article’s lifecycle. The system instantly matches a video when an article goes live, but it also updates the placement if a better, fresher video becomes available later. The editors are empowered with the supervision tools: they retain full control to review, swap, or remove videos after publication, while the system handles the heavy lifting of discovery and continuous optimization.</p><p>Until late 2019, our attempts to solve this relied on integrating various third-party keyword extraction tools and simple category-based matching (e.g., matching a “Sport” video to a “Sport” article). These approaches often failed our internal editorial checks, lacking the nuance required to distinguish between general topic coverage and specific, breaking news events.</p><p>In early 2020, our engineering team shifted strategy. We moved away from external vendors and executed a rapid prototyping phase to build a proprietary Contextual Matching engine. This initial iteration, launched in March 2020, was a keyword-based system tailored for the DACH region. It significantly improved matching quality and eliminated our dependency on external black-box solutions.</p><p>Between 2020 and 2024, our Contextual Player facilitated over <strong>2 billion video views.</strong></p><p>The NLP landscape has shifted largely toward <strong>Transformer architectures.</strong> In 2024, we began upgrading our solution to utilize <strong>semantic retrieval</strong> and <strong>Vector Embeddings. </strong>This shift allows us to understand the intent and context of an article rather than just matching string tokens, reducing configuration overhead and unlocking easy expansion into new languages like Greek, Italian, and Spanish.</p><p>In this article, we will move beyond the business metrics and strictly into the engineering. We will review the algorithms and architecture that powered our growth, detailing how we built a robust keyword engine and how we are now re-architecting for the semantic future.</p><h3>System Architecture: Balancing Latency and Intelligence</h3><p>Building a contextual matching system for news publishers requires balancing two conflicting requirements: <strong>speed </strong>and <strong>intelligence</strong>. When a user loads an article, the video player must load instantly (low latency), but analyzing an article’s text to find the perfect video match takes time.</p><p>To solve this, we designed an asynchronous, event-driven architecture hosted on AWS, primarily built using PHP microservices.</p><h4><strong>1. Configuration and Setup</strong></h4><p>Before a single request is processed, the system is tailored to the specific needs of the publisher. We configure the solution to respect specific categories, channels, and match thresholds. This includes “negative” filtering — defining exactly what topics or tags should be ignored to ensure brand safety or other preferences. It is also possible to set up fallback playlists, in case no videos are matched.</p><p>Then we provide a specific player code to the publisher, who integrates it to all the pages where Contextual Player should appear via the CMS. As mentioned above, this could be even two players per a page, and they would be synchronized (only one will be shown).</p><h4><strong>2. The Synchronous “Hot Path”</strong></h4><p>The player experience begins the moment it loads on the publisher’s page.</p><p><strong>1. Request:</strong> The player calls our <strong>Integration Service</strong>, providing the integration ID and the Article URL.</p><p><strong>2.</strong> <strong>Cache Check:</strong> The Integration Service communicates with the <strong>Contextual Service</strong>, which immediately checks a Redis cache keyed by the article URL.</p><p><strong>3. Immediate Response: </strong><br> ● <strong>Hit</strong>: If a match exists, it is returned immediately. <br> ● <strong>Miss</strong>: If no match exists (e.g., a brand new article), the service returns a pre-configured fallback playlist or default content. This ensures the player never hangs while waiting for an algorithm. It starts playing valid content immediately</p><h4><strong>3. The Asynchronous Matching Flow</strong></h4><p>If the cache was missed, the “intelligence” loop is triggered in the background without blocking the user’s browser.</p><p><strong>1. Queuing</strong>: The Contextual Service pushes a message to an<strong> Amazon SQS </strong>queue, signaling that a new URL needs analysis.</p><p><strong>2. Scraping</strong>: The <strong>Contextual Scraping Service</strong> consumes this message. It fetches the article HTML, extracts the relevant metadata (title, description, keywords, publish date), cleans the data, and pushes the payload to a processing queue.</p><p><strong>3. Matching (The Worker Layer):</strong> This is where the logic lives. Depending on the configuration, one of our worker services picks up the job: <br> ● <strong>Contextual Worker (Legacy/Standard):</strong> Handles the keyword-based logic (which we will detail in the next section). <br> ●<strong> Contextual Semantic Worker (New):</strong> The evolution of our system. This worker leverages Transformer-based algorithms for semantic understanding. It interacts with a <strong>Categorizer Service</strong> — a set of MLP (Multi-Layer Perceptron) models trained via Amazon SageMaker — to classify content types (e.g., distinguishing “Evergreen” content from “Breaking News”) before performing the retrieval.</p><p><strong>4. Persistence &amp; Analytics: </strong><br> ● <strong>Operational</strong>: The final match is written back to Redis. The next time a user loads this article, they get the computed “Best Match” instantly. <br> ● <strong>Analytical</strong>: All match data is stored in a <strong>Snowflake </strong>database. This serves as the source of truth for our internal analytics and allows publisher editors to audit how the system is performing.</p><p>By decoupling the <em>rendering </em>of the video from the <em>matching </em>of the content, we ensure that complex NLP operations never degrade the user experience. <br>For a visual map of how services interact, see the Diagram:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E0jdrGkTBVotvcjiCTUmhA.jpeg" /><figcaption>Diagram by the authors</figcaption></figure><h3><strong>The Keyword-Based Architecture</strong></h3><p>This architecture, which remains the workhorse of our production environment today, was built on a fundamental philosophy: <strong>precision over recall.</strong></p><p>When we began designing this in 2020, our primary focus was the DACH region. This presented a specific linguistic opportunity: <strong>German compound nouns.</strong> While long words like Bundesausbildungsförderungsgesetz (Federal Training Assistance Act) can be a nightmare for standard tokenizers, for us, they acted as perfect “bridges.” If this specific, unique word appears in an article’s metadata and a video’s metadata, the probability of a relevant match is incredibly high.</p><p>We designed a five-stage pipeline to exploit this: <strong>Scraping, Scoring, Categorization, Pricing, and Matching.</strong></p><h4><strong>1. Scraping: Signal over Noise</strong></h4><p>To ensure low latency and high relevance, we made a deliberate choice not to parse the full body text of every article. We found that full-text analysis often introduced more “noise” than signal, picking up peripheral mentions that weren’t central to the story.</p><p>Instead, we focused on the structural elements where editors place the most important information. When a request hits our queue, the<strong> Scraping Service</strong> fetches the HTML and parses the title, description, headers (h1, h2), keywords, and publishing date. This keeps the payload focused on the core topic.</p><h4><strong>2. Scoring: The “Inverse Search” Trick</strong></h4><p>The core engineering challenge was keyword extraction. We maintained a database of millions of “Anchor Keywords.” Crucially, this list wasn’t just a generic dictionary; it was driven by our <strong>video supply.</strong> Through a daily editorial process, new keywords were added only when new videos entered our system. This ensured efficiency: we never waste cycles searching for topics where we don’t have video inventory.</p><p>To extract these keywords efficiently, we utilized an Elasticsearch “Percolator” approach (Inverse Search):</p><ol><li><strong>The Index: </strong>We indexed our Anchor Keywords as documents in OpenSearch.</li><li><strong>The Query: </strong>We treated the article metadata as the query.</li><li><strong>Shingles: </strong>We utilized Shingles (n-gram token filters) to match multi-word concepts.</li></ol><p>Once the raw keywords were identified, we used <strong>spaCy </strong>for a layer of linguistic refinement. We used it primarily for Part-of-Speech (POS) to filter <strong>Nouns </strong>which we used later in addition to the matched Anchor Keywords to build a retrieval query.</p><h4><strong>3. Categorization: The Rules Engine</strong></h4><p>Context isn’t just about text; it’s about business logic. A “Football” article requires different matching behavior than a “Career” article.</p><p>We built a “Waterfall” template system. Based on the article’s metadata, the system selects a Configuration Rule.</p><p>● <strong>Filters</strong>: Does this publisher want videos older than 24 hours? <br>● <strong>Exclusions</strong>: Are there specific “Blocklist” keywords (e.g., sensitive crime topics)?<br>●<strong> Category Logic:</strong> If the article is categorized as “Football,” the system loads a template that applies a massive boost to <strong>Team Names</strong> and <strong>Player Names. </strong>Conversely, in a “Career” category, generic topic matches might carry more weight.</p><h4><strong>4. “Pricing” the Keywords</strong></h4><p>Once we have the keywords and the rules, we need to assign a “price” (or weight) to them. This step effectively constructs the final query. We implemented a sophisticated weighting algorithm:</p><p>● <strong>Location Weight: </strong>Keywords found in the Title get a 2x boost compared to the Description or Headers. <br>● <strong>Length Weight: </strong>Multi-word phrases (shingles) receive higher scores than single tokens, as they usually carry more specific semantic meaning. <br>● <strong>Entity Weight: </strong>Person names are scored differently based on whether we matched the full name or just the surname.</p><h4><strong>5. Matching: The Retrieval</strong></h4><p>Finally, the system queries the <strong>OpenSearch </strong>video index to find the best asset.</p><p>We eventually <strong>removed the standard BM25 (TF-IDF) scoring</strong> from this stage. BM25 is great for general document retrieval, but for exact contextual matching, it often introduced noise. Instead, we relied on our custom “Pricing” boosts. We constructed a Boolean query where the specific “Anchor Keywords” became weighted should clauses.</p><p>We also applied a custom <strong>Synonym Analyzer</strong> at this stage. If the article contained “UK,” the query would automatically boost videos tagged with “United Kingdom” or “Britain,” ensuring we captured the concept, not just the string.</p><h3>The Semantic Future</h3><p>While we cannot share the full technical blueprint of our latest generation just yet, we can outline the architectural shift that is improving the match quality, enabling our expansion into new languages like Greek, Italian, and Spanish, and also significantly reduces the effort required for configuration.</p><p>As effective as the keyword-based system is, scaling it requires constant maintenance of language-specific synonym lists. To solve this, our new <strong>Semantic architecture</strong> adopts a <strong>hybrid approach:</strong> we use semantic vectors for broad retrieval and (sometimes complex) linguistic cues for precise re-ranking.</p><h4><strong>1. Retrieval: Matching Meaning</strong></h4><p>The core innovation is moving the initial search from specific strings to “meaning.”</p><p>● <strong>Encoding</strong>: We use <strong>Sentence Transformers</strong> to map the article’s text into a high-dimensional vector space. Here, we also use the PCA algorithm to reduce the dimensionality of the embeddings. <br>● <strong>Retrieval</strong>: Instead of complex Boolean keyword queries, we perform a <strong>k-NN (k-Nearest Neighbor)</strong> search in OpenSearch.</p><p>This allows the system to identify conceptually related videos (e.g., matching “Automotive Innovation” to “Car Tech”) even if they don’t share a single keyword, effectively solving the “synonym problem” without manual lists.</p><h4><strong>2. Re-ranking: Enforcing Precision</strong></h4><p>However, vector search can sometimes be too abstract. To ensure high editorial standards, we pair the semantic retrieval with a rigid refinement phase. We extract <strong>Stems </strong>and<strong> Named Entities (NER)</strong> from the article and use them to validate the vector candidates. This ensures that while the topic matches semantically, the facts (specific people, teams, or events) align perfectly.</p><p>In our most recent iterations, we are also experimenting with a <strong>fine-tuned LLM</strong> as a final “judge,” assessing the quality of the match to ensure it feels human-curated. Additionally, we’re training internal Deep Neural Network models that allow to categorize videos and articles to different category types, for better re-ranking.</p><h3><strong>Conclusion</strong></h3><p>Contextual matching is a mature, production-grade capability, yet we refuse to treat it as a static “solved” problem. We started with a prototype in 2020 that relied on rules and Python script. Today, we manage a system that blends rigid keyword logic with fluid semantic vectors, constantly iterating based on our own internal R&amp;D and the latest breakthroughs in NLP.</p><p>For our publishers, the technology is invisible. They simply see that their articles are populated with relevant, high-quality videos that keep their readers engaged. For us engineers, the journey from “keyword matching” to “hybrid semantic understanding” has been the key to unlocking scale — delivering over 2 billion views and counting.</p><p>While we cannot share every proprietary detail or specific implementation nuance in this public article, we hope you found this look under the hood valuable. We do appreciate your time and interest in our technology</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9cfa54ee85e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/from-keywords-to-vectors-how-glomex-built-a-contextual-video-matching-engine-at-scale-a9cfa54ee85e">From Keywords to Vectors: How Glomex Built a Contextual Video Matching Engine at Scale</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automation projects — a real game changer or an expensive failure?]]></title>
            <link>https://medium.com/tech-p7s1/automation-projects-a-real-game-changer-or-an-expensive-failure-272d6494916d?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/272d6494916d</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Wed, 21 May 2025 08:17:34 GMT</pubDate>
            <atom:updated>2025-05-21T08:17:34.138Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Automation projects — a real game changer or an expensive failure?</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*hzicM8zyKdauotZVq54Tng.png" /><figcaption>Picture by <strong>Philipp Slupina</strong></figcaption></figure><p>by<strong>: Philipp Slupina</strong></p><p><strong>Our automation team has developed an Excel template, that makes you stay in control and maximize your chances of success. Read all about the process behind and get the Excel table.</strong></p><h3>Table of Contents</h3><p>· <a href="#082b">Ownership</a><br>· <a href="#9f69">Early-Stage Implementation</a><a href="https://medium.com/tech-p7s1/aws-account-management-closing-04765f0e804a#46f0"><br></a>· <a href="#1bf3">Benefits of a Structured Evaluation Approach</a><br>· <a href="#dccc">Theoretical Foundations and Alternative Approaches</a><br>· <a href="#d799">Case Study: Automation of Invoice Processing</a><br>· <a href="#9cf5">Features of the Excel Template</a><br>· <a href="#e061">Addressing Challenges and Concerns</a></p><p>In this blog post, we’re going to take a closer look at project evaluation in automation. You probably already know that the main goals of automation are to boost efficiency, cut costs, and make the best use of resources. But here’s the thing: not every automation project ends up delivering the cost savings or productivity improvements you might expect. That’s why doing a thorough evaluation early on is so important. It helps figure out if the project is feasible and ensures you’re making smart investment choices.</p><p>As a team committed to diving into the latest in automation, we’ve put together a handy Excel-based calculation model to help you estimate how profitable your project could be. This tool makes decision-making a lot easier by letting you compare costs and benefits side by side. That way, organizations can make choices based on solid data. And good news! You can download the Excel <strong>template</strong> and tweak it as needed, so it fits your specific projects perfectly.</p><p><strong>But before you rush to plug in your numbers, there’s one critical factor that even experienced teams often overlook — one that could make or break your entire project.</strong></p><h4><strong>Ownership</strong></h4><p>In the context of automation, ownership means clearly defining <em>who</em> is responsible for what — before, during, and after the project is implemented. It’s not just about assigning tasks; it’s about accountability. Who ensures the data feeding into the system is accurate? Who handles updates and maintenance? Who steps in when something breaks? Without clear ownership, even the best-planned automation solutions can falter due to confusion, finger-pointing, or simple neglect.</p><p>Establishing ownership early on helps keep your project aligned with business goals, ensures smoother implementation, and builds long-term sustainability. It’s the glue that holds your strategy together.</p><h4><strong>Early-Stage Implementation</strong></h4><p>This template is intended for use at the early stages of a project — typically after initial discussions with key stakeholders. At this point, sufficient data is available for an initial profitability assessment. By leveraging this tool early, you can determine whether further investment is justified or if alternative approaches should be considered.</p><p><strong>Typical Steps of an Automation Project</strong></p><p>Automation projects generally follow a structured process to ensure successful implementation. The key steps include:</p><ol><li><strong>Identification of Automation Opportunities:</strong></li></ol><ul><li>Assess existing workflows and identify repetitive, time-consuming, and error-prone tasks.</li><li>Conduct stakeholder discussions to understand pain points and potential improvements.</li></ul><p><strong>2. Feasibility Analysis and Business Case Development:</strong></p><ul><li>Evaluate technical and financial feasibility, including expected ROI.</li><li>Develop a business case with cost-benefit analysis and risk assessment.</li></ul><p><strong>3. Solution Design and Requirement Gathering:</strong></p><ul><li>Define functional and technical requirements.</li><li>Choose suitable automation technologies such as RPA, AI, or workflow automation tools.</li></ul><p><strong>4. Proof of Concept (PoC) and Pilot Implementation:</strong></p><ul><li>Develop a small-scale pilot to validate feasibility.</li><li>Gather feedback and make necessary adjustments before full deployment.</li></ul><p><strong>5. Full-Scale Implementation:</strong></p><ul><li>Develop and deploy the automation solution.</li><li>Train employees and integrate automation into existing processes.</li></ul><p><strong>6. Monitoring and Optimization:</strong></p><ul><li>Continuously track performance against expected benefits.</li><li>Optimize workflows and address any emerging issues.</li></ul><p><strong>7. Scaling and Expansion:</strong></p><ul><li>If successful, extend automation to other areas of the business.</li><li>Ensure long-term sustainability through ongoing improvements and maintenance.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/564/1*eUhAzrTpU6czaDpxOq8r1g.png" /><figcaption>Picture by <strong>Philipp Slupina</strong></figcaption></figure><h4><strong>Benefits of a Structured Evaluation Approach</strong></h4><p>A systematic approach to project evaluation offers multiple advantages:</p><ol><li><strong>Objective Decision-Making:</strong> By standardizing data collection, the evaluation process becomes independent of subjective opinions, resulting in objective, data-driven decisions.</li><li><strong>Enhanced Transparency:</strong> Clear visualization of costs and anticipated savings fosters transparency, making financial impacts understandable for all stakeholders and facilitating internal communication.</li><li><strong>Early Identification of Risks and Opportunities:</strong> A detailed cost-benefit analysis helps in identifying potential risks and implementing mitigation strategies while uncovering untapped opportunities that can further enhance project value.</li><li><strong>Prioritization of High-Value Projects:</strong> Not all automation initiatives yield the same economic benefits. This template enables organizations to rank projects based on profitability, ensuring that the most valuable initiatives receive priority.</li><li><strong>Comprehensive Documentation:</strong> Beyond individual projects, the accumulated documentation offers insights into the overall value contributed by the automation team, demonstrating the long-term impact of automation efforts.</li></ol><h4><strong>Theoretical Foundations and Alternative Approaches</strong></h4><p>While this template focuses on ROI as a primary evaluation metric, other established financial and strategic methodologies can provide a more comprehensive assessment of automation projects:</p><ul><li><strong>Return on Investment (ROI):</strong> Measures the financial return of an investment compared to its cost. A high ROI indicates strong financial justification, while a low ROI may require reconsideration.</li><li><em>Formula:</em> ROI = (Net Profit / Cost of Investment) × 100</li><li>ROI provides a straightforward metric for comparing different automation projects. (<a href="https://online.hbs.edu/blog/post/how-to-calculate-roi-for-a-project">source</a>).</li><li>Variations of ROI include <strong>Annualized ROI</strong>, which adjusts for time, and <strong>Risk-Adjusted ROI</strong>, which factors in potential uncertainties.</li><li><strong>Net Present Value (NPV):</strong> Assesses the present value of an investment to determine long-term profitability.</li><li><strong>Internal Rate of Return (IRR):</strong> Measures the project’s rate of return, facilitating comparisons with other investments.</li><li><strong>Payback Period:</strong> Determines the duration required to recover the initial investment.</li><li><strong>Total Cost of Ownership (TCO):</strong> Accounts for both direct and ongoing maintenance costs.</li><li><strong>Balanced Scorecard (BSC):</strong> A strategic planning method that evaluates automation projects based on financial and non-financial perspectives, including customer satisfaction, internal processes, and learning/growth opportunities.</li><li><strong>Risk-Adjusted Return (RAR):</strong> Integrates potential risks into the return calculations to provide a more realistic profitability assessment.</li><li><strong>Real Options Analysis:</strong> Examines the flexibility of investment decisions in automation, enabling companies to adapt to changing circumstances.</li></ul><p>Don’t just take our word for it! Several authoritative studies and reports from organizations such as McKinsey, Deloitte, and the Fraunhofer Institute further underscore the effectiveness of automation in driving cost efficiency and productivity gains. For instance, McKinsey highlights that automation can significantly reduce operational costs (<a href="https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/how-industrial-companies-can-cut-their-indirect-costs-fast">source</a>). Deloitte also emphasizes the importance of accurate ROI measurement in automation (<a href="https://www2.deloitte.com/us/en/pages/advisory/articles/measuring-enterprise-automation-roi.html">source</a>).</p><h4><strong>Case Study: Automation of Invoice Processing</strong></h4><p><strong>Scenario:</strong></p><p>A mid-sized company manually processes 5,000 incoming invoices monthly, with each invoice requiring approximately 10 minutes of processing time. This results in significant labor costs and a high likelihood of errors.</p><p><strong>Solution:</strong></p><p>Implementation of an automated invoice processing system utilizing OCR technology and workflow integration.</p><p><strong>Expected Benefits:</strong></p><ul><li>Reduction in processing time per invoice from 10 minutes to 2 minutes.</li><li>Monthly savings of 667 working hours.</li><li>Estimated ROI over three years: <strong>73%</strong>.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*7_hLn6N8ODxuTfVGFKz8Fw.png" /><figcaption><a href="https://onedrive.live.com/?redeem=aHR0cHM6Ly8xZHJ2Lm1zL2YvYy8yYTQ5ZjQ2ZmIyNjA4ZjljL0VzZ2wzWjZhZy1wSnNXdEtPbHd6eXVRQnlhRHl2bXRwdU9VR3E1RUo5U01GUnc%5FZT1aTHY2UFg&amp;id=2A49F46FB2608F9C%21s9edd25c8839a49eab16b4a3a5c33cae4&amp;cid=2A49F46FB2608F9C">Excel Template</a> made by <strong>Philipp Slupina</strong></figcaption></figure><h4><strong>Features of the Excel Template</strong></h4><p>The Excel template is designed to facilitate thorough project evaluation by incorporating the following components:</p><ul><li><strong>Project Cost Documentation:</strong> Track one-time and recurring costs.</li><li><strong>Benefit Analysis:</strong> Estimates expected savings in financial terms and work hours.</li><li><strong>Automated Calculations:</strong> Compares costs and benefits dynamically.</li><li><strong>Three-Year Profitability Assessment:</strong> Provides a long-term financial perspective.</li><li><strong>ROI Calculation:</strong> Offers a quick and reliable profitability assessment.</li><li><strong>Operational Cost Share Analysis:</strong> Determines the percentage of project expenditures allocated to ongoing operations.</li><li><strong>Estimated Employee Cost Rates:</strong> Integrates realistic daily rates for IT and business teams.</li></ul><h4><strong>Why 3 years?</strong></h4><p>Software automation solutions typically have a lifespan of <strong>5 to 10 years</strong>, depending on factors like the type of solution, industry, and technological advancements. While some solutions, particularly in industries like manufacturing, may last longer, sectors with rapid changes, like software development, may need more frequent updates.</p><p>Regarding depreciation, standard software is typically written off over <strong>3 years</strong> in countries like Germany, with custom solutions sometimes extending to <strong>5 years</strong>. This period reflects the fast-paced nature of technological innovation. In my opinion, the <strong>3-year depreciation</strong> period is beneficial for businesses, as it aligns with the frequent updates and technological shifts common in today’s market. It allows companies to quickly account for software costs and reinvest in newer, more efficient solutions without being burdened by outdated systems. This approach ensures financial flexibility while keeping up with the latest advancements.</p><h4><strong>Addressing Challenges and Concerns</strong></h4><p>While an ROI template enhances transparency, it may also raise concerns. The following challenges should be considered:</p><ul><li><strong>Quantifying Intangible Benefits:</strong> Certain advantages, such as improved customer and employee satisfaction, may be difficult to express in monetary terms.</li><li><strong>Resistance to Change:</strong> Employees may fear job displacement due to automation. Transparent communication regarding the optimization goals is crucial.</li><li><strong>Long-Term Maintenance Costs:</strong> Beyond initial investments, ongoing expenses for updates and support must be factored in.</li><li><strong>Accounting for Qualitative Factors:</strong> Some automation benefits, such as scalability and improved data quality, may not be immediately reflected in financial figures.</li></ul><h4><strong>Conclusion: A Strategic Tool for Smarter Investment Decisions</strong></h4><p>This Excel template serves as more than just a financial calculation tool — it is a strategic instrument for evaluating automation projects. By facilitating informed decision-making, increasing transparency, and optimizing resource allocation, it empowers organizations to drive automation initiatives with confidence.</p><h4><strong>About the Author</strong></h4><p><a href="https://www.linkedin.com/in/pslupina/"><strong>Philipp Slupina</strong></a><strong> </strong>is a seasoned automation expert with a track record of successfully implementing process automation strategies. With extensive experience in evaluating and executing automation projects, Philipp helps maximize efficiency and ROI through structured methodologies. Connect with him on LinkedIn for further discussions on automation strategies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/368/1*3vH9e5OwkRFP6LRSMVCovw.png" /><figcaption><a href="https://www.linkedin.com/in/pslupina/"><strong>Philipp Slupina</strong></a><strong> </strong>(Head of ITSM &amp; Automation Process Automation / ITSM)</figcaption></figure><blockquote><strong><em>“I am excited to share this template with professionals interested in automation and welcome feedback for further enhancements!”</em></strong></blockquote><p><strong>Ready to take control of your automation projects?</strong></p><p>Download<strong> </strong><a href="https://onedrive.live.com/?redeem=aHR0cHM6Ly8xZHJ2Lm1zL2YvYy8yYTQ5ZjQ2ZmIyNjA4ZjljL0VzZ2wzWjZhZy1wSnNXdEtPbHd6eXVRQnlhRHl2bXRwdU9VR3E1RUo5U01GUnc%5FZT1aTHY2UFg&amp;id=2A49F46FB2608F9C%21s9edd25c8839a49eab16b4a3a5c33cae4&amp;cid=2A49F46FB2608F9C"><strong>the Excel template</strong></a><strong> </strong>and start maximizing your success!</p><p>(Are you having trouble with the template? Be sure to<strong> download the file</strong> instead of just clicking on the link to view it online. Opening it directly in Excel is the best way to guarantee everything works as intended.)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=272d6494916d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/automation-projects-a-real-game-changer-or-an-expensive-failure-272d6494916d">Automation projects — a real game changer or an expensive failure?</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Celebrating one year 7.1 Tech Hub]]></title>
            <link>https://medium.com/tech-p7s1/celebrating-one-year-7-1-tech-hub-52a4db3876bb?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/52a4db3876bb</guid>
            <category><![CDATA[tech-hub]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Thu, 09 Jan 2025 13:53:43 GMT</pubDate>
            <atom:updated>2025-01-09T13:53:43.217Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VvdmEyK5XUv6SS2P6rLg6w.jpeg" /><figcaption><strong>Our team at 7.1 Tech Hub</strong></figcaption></figure><p>About a year ago, P7S1 launched the 7.1 Tech Hub in Porto/Portugal.The Tech Hub complements our existing offices in Munich/Unterfoehring and Berlin, providing us with a great location to attract Tech and Data talent across various domains.</p><p>A lot has happened in the past 365 days. What started out with a founding team of three people has emerged into a work force of around 40 employees. And there is plenty more the team has achieved in the past year.</p><p><strong>Team growth &amp; new office location</strong> 🎉</p><p>The 7.1 Tech Hub now consists of a team of around 40 employees from five different countries. From Software Developers, Data Scientists and BI Engineers — the knowledge of our Portuguese, Brazilian, German, Italian and Polish colleagues result in fresh ideas and innovative solutions that bring our Entertainment business forward. From advancing the technology behind our streaming platform Joyn to enhancing the internal use of AI and working hand in hand with our internal IT service provider PTS (ProSiebenSat.1 Tech &amp; Services GmbH). And also the 7.1 Tech Hub Management Team was extended by two Portuguese colleagues (Pedro &amp; Zé).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tWq24amjApCa1Htj2iWFdQ.jpeg" /><figcaption>From left to right: Richard Seitz (Co-founder &amp; former Managing Director), Pedro Fialho (HR Country Lead), Steffi Beckmann (Co-founder &amp; Managing Director), Lucas Braun (Co-founder &amp; Head of Data &amp; AI), Zé Carvalho (Head of IT)</figcaption></figure><p>To offer our employees a modern working environment and to drive the operational business forward, the team has furthermore moved its office location.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3fL3xlUO58ym2OIcJkoFVQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*plWx_mNPfRkvCwXfIMa_Ww.jpeg" /><figcaption>Setting up the new office</figcaption></figure><h4>First important milestones</h4><p>After less than a year, the 7.1 Tech Hub was recognized as the best Tech employer and the fastest growing Tech company in Portugal, even ahead of established industry players. This great success is based on the consistently positive reviews from candidates and employees and shows the hard work, commitment and passion of the entire team.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j-s6QiQmhYIUlkAUmjCNLg.png" /><figcaption>Stefanie Beckmann, Co-founder &amp; M</figcaption></figure><blockquote>“It’s hard to believe that we only founded the Tech Hub a year ago! What we’ve achieved in such a short time is simply fantastic — and that’s mainly down to the strong team performance of the management team and our colleagues from Germany. Many thanks to all of them! Of course, I am particularly proud of our team here in Porto: Not only are they driving Tech and Data Solutions to the next level together, they are also bringing innovative ideas to our processes and working methods. And this is just the beginning — we still have big plans!”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/1*CYwjAQBcvPXJ7V23oHjIyQ.jpeg" /><figcaption><em>1st place as fastest growing company in the Teamlyzer ranking (Portuguese employer evaluation platform)</em></figcaption></figure><p>We look back on a successful year full of growth, team spirit and innovation. We are excited to see what the new year has in store for our team in Porto and what successes we will soon be able to share with you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PYE8O8fc5Zbu2p8BA02KdA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MtbIgtfJL_7CmqkFcjpVJw.jpeg" /><figcaption>left: PTS CTO Wolfgang Neubauer visiting the Tech Hub and sharing the strategy; right: Halloween celebrations in the office</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=52a4db3876bb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/celebrating-one-year-7-1-tech-hub-52a4db3876bb">Celebrating one year 7.1 Tech Hub</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[KARGO (pt 5) — Managing custom EKS cluster addons with Flux CD]]></title>
            <link>https://medium.com/tech-p7s1/kargo-pt-5-managing-custom-eks-cluster-addons-with-flux-cd-0a9b6a706088?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/0a9b6a706088</guid>
            <category><![CDATA[eks-cluster]]></category>
            <category><![CDATA[configuration-management]]></category>
            <category><![CDATA[kubernetes-addons]]></category>
            <category><![CDATA[flux-cd]]></category>
            <category><![CDATA[helm-chart]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Thu, 12 Dec 2024 15:13:37 GMT</pubDate>
            <atom:updated>2025-01-16T16:27:35.099Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>KARGO (pt 5) — Managing custom EKS cluster addons with Flux CD</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YVgV7V9KwCF6WQ72pufK6A.jpeg" /></figure><h3><strong>Audience</strong></h3><p>In this article, we will describe why we use Flux CD to deploy tooling on EKS clusters. We also scratch the surface on how flux works and how we build a workflow around Flux CD to make sure we don’t break too many things.</p><h3><strong>Problem</strong></h3><p>As you saw in the overall architecture <a href="https://medium.com/tech-p7s1/moving-up-the-stack-c680cebe234c">article</a>, we provide our customers not an empty EKS cluster, but a cluster with preinstalled addons which make the life of our customers easier.</p><p>Managing several clusters by hand takes up a lot of your time and is prone to configuration drift.</p><p>So we need a way to automatically install and manage software on multiple EKS clusters so that all clusters are configured the same way and have the same versions installed. The clusters have to become cattle and not pets. Read more about cattle vs pets in this <a href="https://medium.com/tech-p7s1/kargo-a-container-focused-developer-platform-on-aws-0bdc5262fa46">article.</a></p><h3><strong>Solutions</strong></h3><p>The same problem was experienced by other people and two popular solutions were created in the open source tool landscape.</p><ul><li>Argo CD</li><li>Flux CD</li></ul><p>We tested both and came to the conclusion that we want to use Flux CD.</p><h3><strong>Why Flux CD</strong></h3><p>Before we moved to the cloud we already deployed all our software with <a href="https://helm.sh/">Helm</a> and Flux has the feature to deploy Helm charts, so no big changes here. In addition Flux can also deploy plain resource manifests, so all options are available to us.</p><p>With Argo CD we would need to establish a new workflow where we render the helm charts for Argo CD to deploy them.</p><p>You can also check out a comparison between Argo and Flux <a href="https://blog.aenix.io/argo-cd-vs-flux-cd-7b1d67a246ca">in this article.</a></p><h3><strong>Cluster ResourceSets for Flux installation</strong></h3><p>Ok Flux it is, but how do you install it?</p><p>Easy, with Argo CD …just kidding.</p><p>As we create the EKS cluster with ClusterAPI, we can use <a href="https://cluster-api.sigs.k8s.io/tasks/experimental-features/cluster-resource-set">ClusterResourceSets</a> to install Kubernetes objects on the newly created cluster.</p><p>You can also check the previous article about <a href="https://medium.com/tech-p7s1/kargo-series-part-3-create-kubernetes-clusters-at-scale-with-cluster-api-eff259d06874">ClusterAPI</a></p><p>The time this post was written we are installing the following resources with ClusterResourceSets:</p><ul><li>Namespaces</li><li>NetworkPolicies for Kubernetes Addons (additional software needed on every EKS cluster) as they are installed via EKS magic and not a Helm Chart.</li><li>Configmaps for tools which need some cluster specific values (e.g. IP ranges, DNS Domain)</li><li>Flux</li></ul><h3><strong>Design/Architecture</strong></h3><p>The Design was inspired form <a href="https://github.com/fluxcd/flux2-multi-tenancy">Flux Repository examples.</a></p><h3><strong>Repository structure</strong></h3><p>Currently we have two “flavours” of clusters: dev and prod</p><ul><li>The base directory contains HelmRepository and HelmRelease manifests</li><li>The dev directory contains HelmRelease Kustomize patches for deploying pre-releases on the dev cluster</li><li>The prod directory contains HelmRelease Kustomize patches for deploying stable releases on the production cluster</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/417/1*hl0wYBCOXcrbGNqhjr73yA.png" /></figure><h3><strong>How does Flux work</strong></h3><p>When bootstrapping Flux, besides the deployments of the Flux tools itself (helm-controller, kustomize-controller, notification-controller, source-controller), a “Secret”, a “GitRepository” and a “Kustomization” object is created which tells Flux to which git repository and which branch to connect to.</p><p>Next you can find the GitRepository Object. Important parts are the “URL” of your GitRepository, the “branch” of your GitRepository flux has to use (this will be handy when testing new releases of tools installed by flux) and the “secretRef” which is a secret containing a token for your git repository.</p><p>❯ kubectl get gitrepository flux-system -n flux-system -oyaml</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/1*jazvSihgtEVJPaMh-GiguQ.png" /></figure><p>Here is an example of the “Kustomization” object telling Flux which path/directory in your git repository to use. This is used to differentiate clusters between “dev” and “prod”.</p><p>❯ kubectl get gitrepository flux-system -n flux-system -oyaml</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/589/1*BRz7hq9N6uO3CF_DFLYs7Q.png" /></figure><p>Flux then reads the relevant directories and files. In this example, Flux connects to the directory clusters/dev. There it reads the 4 yamls (apps.yaml, infrastructure-basics.yaml, infrastructure-crds.yaml, infrastructure.yaml)</p><p>We concentrate on “infrastructure.yaml” and how “cert-manager” is installed. This is a shortened example.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*uHPpKvV_8KDR4pDvTT76yw.png" /></figure><p>Step 1, infrastructure.yaml. This tell Flux to go to “infrastructure/dev”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/571/1*kGzmUQktRd4ANLNNUVO1dA.png" /></figure><p>Step 2, kustomization.yaml, as “infrastructure/dev” is a directory, Flux looks for a file named “kustomization.yaml”, in this file other directories/resources are defined. We concentrate on “cert-manager”, but the same happens to all resources.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/664/1*EaK0vJW1-lllVn9dJyULuQ.png" /></figure><p>Step 3, kustomization.yaml in the cert-manager directory. Again, as “cert-manager” is a directory Flux looks for the “kustomization.yaml”. In this file we tell Flux to look for resources in “../../base/cert-manager” directory, but also apply a patch to this resources from the file “helmrelease-patch.yaml”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/565/1*nuitSmV_SWsjj2Gaxywt-Q.png" /></figure><p>Step 4, helmrelease-patch.yaml. This tells Flux to override the helm chart version of the cert-manager (form step 6). This is handy to run a different version in the dev cluster compared to the prod cluster.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/596/1*u6FQODi-dnujyUl-kO5b2A.png" /></figure><p>Step 5, kustomization.yaml in the cert-manager directory. Again, as “cert-manager” is a directory Flux looks for the “kustomization.yaml”. This tells Flux to look for a file/resource named “helmrelease.yaml”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/568/1*x62_tL5HIds--WJlGCF0gA.png" /></figure><p>Step 6, helmrelease.yaml. And finally we tell Flux what to install. You can see we define values files, and also using a ConfigMap to build up the needed values fo Helm. The ConfigMap “cert-manager-values” was placed in the kube-system namespace by the ClusterAPI using ClusterResourceSets. As you see, we do not define any version. The version is patched in from “Step 4”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/564/1*ZXSECb786k1GOkCSVOMJVw.png" /></figure><h3><strong>Dependencies</strong></h3><p>Dependencies can become tricky as you cannot define dependencies between all objects.</p><p>As far as I understand you can define dependencies between Helm charts or between Kustomizations but not between Helm charts and Kustomizations.</p><p>As an example, in “Step 1” you can see that “infrastructure” is depended on “infrastructure-crds”. Both are “Kustomizations”. So this way we force Flux to install all crds first, and then everything which is in “infrastructure”</p><p>In “Step 6” we have a dependency on “aws-load-balancer-controller” which is a Helm dependency. Flux will first install the “aws-load-balancer-controller” before installing the “cert-manager”.</p><h3><strong>Our Workflow</strong></h3><p>It took us some time to come up with a workflow which is not too elaborate but also not too prone to brake something. Also we don’t have any fancy testing frameworks in place (yet), which would simplify the workflow, so we have to do the testing by hand.</p><p>If we want to add or update something installed by Flux, we use the following Steps:</p><ol><li>Create a new branch in the repo Flux is reading.</li><li>Spin up a test cluster and connect it to the branch (k edit gitrepository flux-system -n flux-system) (test clusters already use the “clusters/dev” path)</li><li>Make changes in the “dev” directories (e.g. change the version of a helm chart “See Step 4”)</li><li>Push changes</li><li>Check what Flux is doing (“k get helmrelease -A”, “helm ls”, “k get pods”, …)</li><li>If all is good, let someone merge it. This will cause the dev cluster (which is always running, and attached to the “main” branch) to install the new versions.</li><li>Check if the dev cluster is also OK.</li><li>Create a new branch as a followup to the previous branch, and redo what you did to the “dev” part in the “prod” part.</li><li>Let someone merge it, it will now be rolled out on all “prod” clusters.</li></ol><h3><strong>Tipps</strong></h3><p>To see what is happening you can use flux get all -A, this will show all the resources Flux is using and their status.</p><p>When something takes too long, or you fat fingered (make a typo) something up in the Helm chart, you can use some Brute Force IT™. Just delete the flux pods which will force Flux to recheck everything. Also as the source-controller acts as a source for the Helm charts, this forces a redownload of all Helm charts from the registry. The Helm charts are pulled from the registry, repacked, and made downloadable in the cluster. This repacked version is then installed by the helm-controller.</p><p>If you have some changes in the Helm chart that cannot be upgraded (e.g. changes in matchlabels section), you can instruct flux to uninstall the Helm release on failure, and do a fresh install. But this should only be done if you exactly know what will happen, having this on by default, could result in a unfavorable behaviour. For example you are trying to install a new version but the new version is broken and will never work then this would uninstall the working version and you end up with nothing running. Check “Step 6” for the right place for this.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/1*X1QtJDEvK1Y6s282Mc7ilg.png" /></figure><h3><strong>Conclusion</strong></h3><p>Flux CD assures that all EKS clusters are configured the same way and there is no configuration drift. We can deploy the tools with helm charts, which we already had. We can also deploy plain resources if a need arises. The possibility to define dependencies helps preventing installation or upgrade errors. It provides an easy way to manage a fleet of clusters in a single repository. We can easily divide the clusters into more flavours when the necessity arises.</p><h3>More about KARGO</h3><ul><li><a href="https://medium.com/tech-p7s1/moving-up-the-stack-c680cebe234c">KARGO (pt 1) — Moving up the stack</a></li><li><a href="https://medium.com/tech-p7s1/kargo-a-container-focused-developer-platform-on-aws-0bdc5262fa46">KARGO (pt 2) — A container focused developer platform on AWS</a></li><li><a href="https://medium.com/tech-p7s1/exposing-workloads-on-eks-0ec39acd5fa9">KARGO (pt 3) — Exposing workloads on EKS</a></li><li><a href="https://medium.com/tech-p7s1/kargo-series-part-3-create-kubernetes-clusters-at-scale-with-cluster-api-eff259d06874">KARGO (pt 4) — Create Kubernetes cluster at scale with ClusterAPI</a></li><li><strong>KARGO (pt 5) — Managing custom EKS cluster addons with Flux CD</strong></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0a9b6a706088" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/kargo-pt-5-managing-custom-eks-cluster-addons-with-flux-cd-0a9b6a706088">KARGO (pt 5) — Managing custom EKS cluster addons with Flux CD</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[P7S1 x Munich Datageeks]]></title>
            <link>https://medium.com/tech-p7s1/p7s1-x-munich-datageeks-679e7ba42b25?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/679e7ba42b25</guid>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Wed, 04 Dec 2024 09:56:55 GMT</pubDate>
            <atom:updated>2024-12-04T09:56:55.239Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CDxlNh7u6jti1yUB3OFtSQ.png" /></figure><p><em>Last week, in an exciting collaboration with the non-profit organization Munich Datageeks, we hosted a pre-Christmas Meetup at our campus in Unterföhring for the second time, diving deep into the fascinating worlds of Data and AI.</em></p><p>The heart of the Datageeks community lies in its mission to connect passionate individuals who aspire to make a meaningful impact in the digital arena while facilitating a rich exchange of knowledge. Also this time, our Meetup was a great opportunity for attendees to share their insights, spark innovative discussions, and explore the dynamic universe of our P7S1 Tech &amp; Data ecosystem.</p><p>We opened our doors to a diverse crowd — students, Academic Researchers, Software Engineers, Statisticians, Data Scientists, and Entrepreneurs — all eager to network, collaborate, and inspire one another. Together, we ignited a lively atmosphere filled with curiosity and enthusiasm as everyone delved into the fascinating possibilities that AI offers in regard of planning optimization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/644/1*4kLcUExM1BWZN43jd2SBkQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/680/1*33ffbxczZcdJ4wkmw1O-Eg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/646/1*HGPbNVSwDzJApc7QtW__9w.png" /><figcaption>picture left: Mark (Verivox) &amp; Torsten (Data Geeks) left to right; middle picture: Hanna, P7S1; picture right: Violetta, Aleksandar, Nilabhra, Hanna (all P7S1) left to right</figcaption></figure><h4><strong>The evening’s insights in a nutshell</strong></h4><p>The evening commenced with an introductory session by Torsten Schön, from Munich Datageeks, and Richard Seitz, SVP Data Products, leading into insightful discussions about AI and data.</p><p>Our colleagues — Aleksandar (Data Architect), Violetta (Business Intelligence Engineer), Nilabhra (Machine Learning Engineer), and Hanna (Associate Product Manager) — articulated how P7S1 leverages AI to refine its television planning decisions.</p><p>This strategic approach has allowed us to obtain data-driven insights that help strengthening our market share in movies and shows, even in the face of limited planning time. The importance of collaboration among the Data Platforms, Business Intelligence, and AI teams emerged as a critical theme in achieving these objectives. In this regard, our experts first provided a detailed overview of the unified data lake, illustrating its role in data provision and the underlying infrastructure that supports it. Attendees also gained valuable insights into data integration processes and the organization of KPI calculations. Furthermore, we delved into the methodologies employed by AI, utilizing AWS and other tools to forecast market shares effectively. To conclude the session, we were shown an overview of how these predictive analytics are visually represented for planners, thereby facilitating informed decision-making in their strategic planning efforts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/445/1*kHtFPhK6ICSCXbiBhYOMLw.png" /><figcaption>Hanna, Associate Product Manager</figcaption></figure><blockquote>“Collaboration among the Data Platforms, BI, and AI teams has been the cornerstone of our success in delivering significant stakeholder value. By turning data into actionable insights, we empower our stakeholders to make informed program scheduling decisions that drive impactful outcomes.”</blockquote><p>In the second segment of the event, Hellmar, Senior Sales Engineer at Imply, delivered an in-depth presentation on Apache Druid, a high-performance, distributed real-time analytics database. This database is used particularly for streaming analytics applications. Hellmar highlighted how Apache Druid effectively analyzes streaming data, demonstrating its numerous advantages, such as high concurrency at minimal costs, the ability to leverage both historical and real-time insights, and automated fault tolerance. These features make it an invaluable tool for organizations seeking efficient data processing solutions. Following this overview, Hellmar presented a practical use case involving publisher clickstream data. This example illustrated the database’s capabilities in real-world scenarios, further displaying its potential to enhance data-driven decision-making.</p><h4><strong>Time for exchanging ideas and networking</strong></h4><p>Once the presentations wrapped up and the valuable takeaways were shared, the focus shifted to networking in a relaxed and welcoming atmosphere. During a delightful get-together, guests had ample opportunity to mingle, exchange ideas, and forge new connections over food and drinks, including some delicious pizza that added to the evening’s enjoyment. We extend our heartfelt thanks to the Munich Datageeks and all attendees for contributing to such an insightful event and fostering a spirit of collaboration and knowledge sharing. We’re looking forward to the next time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/773/1*LKsB9ABbJoJ4XBJyJ1jiYQ.png" /><figcaption>Julia and Markus (both P7S1)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=679e7ba42b25" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/p7s1-x-munich-datageeks-679e7ba42b25">P7S1 x Munich Datageeks</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[By Aleksandar Culafic​]]></title>
            <link>https://medium.com/tech-p7s1/by-aleksandar-culafic-91af1a5b5f02?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/91af1a5b5f02</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[power-automate]]></category>
            <category><![CDATA[greeting-bot]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Tue, 29 Oct 2024 16:07:44 GMT</pubDate>
            <atom:updated>2024-10-29T16:07:44.818Z</atom:updated>
            <content:encoded><![CDATA[<p>By Aleksandar Culafic​</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E1w_WrziGt7hdSZ6UdiNAw.jpeg" /></figure><h3>How to Greeting Bot</h3><p>Automating the onboarding process in organizations can significantly improve efficiency and team dynamics. A greeting bot, developed with Power Automate, introduces new team members in Microsoft Teams channels on a regular and structured basis.</p><p><strong>What is Power Automate?</strong></p><p>“Power Automate” is a tool for automating workflows, focused on avoiding repetitive tasks and optimizing processes. It is part of the “Power Platform,” a comprehensive platform made up of several tools that work together to meet various business requirements. The bot sends automated messages on the 1st and 15th of each month to introduce new colleagues. These messages include names, departments and job titles. These personalized greetings ensure that the entire team is informed about new members and that they feel welcomed right from the start.</p><p><strong>Benefits of Using Power Automate</strong></p><p>Power Automate allows to create custom workflows without in-depth programming knowledge. The flow is set to send messages only on specified days, excluding weekends or holidays. This increases the possibility that team members will see and engage with the messages.</p><p>The greeting bot significantly reduces the administrative workload for team administrators by automating manual tasks and ensuring consistent communication. This not only increases efficiency, but also promotes a friendly and welcoming atmosphere in the team. New members feel welcome thanks to the personal approach. Overall, the bot strengthens the team dynamics, improves internal communication, and saves time.</p><p><strong>Creating the Welcome Message</strong></p><p>When drafting the welcome message, several requirements must be met:</p><p>· <strong>Clear structure:</strong> the message should be clear and well-organized for easy readability</p><p>· <strong>Personalization: </strong>Include personalized information, such as names, departments and working titles of new members</p><p>· <strong>Relevant content: </strong>It should list all relevant details about new employee onboarding that are of interest to the team</p><p>· <strong>Adaptability: </strong>The message should align with the company’s specific needs</p><p>· <strong>Courtesy and professionalism: </strong>Maintain a friendly and professional tone</p><p>· <strong>Visual design: </strong>Use appealing visual elements or formatting where appropriate to increase reader attention</p><h3>Who we are and what we do</h3><p>We are the process automation team and act as an internal <strong>Center of Excellence (CoE)</strong> that establishes and develops process automation within the Group. We aim to build a <strong>citizen developer community</strong> in order to integrate automation directly into the teams.</p><p>As a point of contact, we assist in creating apps and workflows. For complex projects with high return on investment potential, we also offer <strong>full development services</strong>. In such cases, we go through the entire process step by step together with the contact person. We <strong>document, scrutinize, and analyze </strong>the individual steps to determine whether and how automation can be implemented sensibly and profitably. Our approach ensures a thorough and targeted implementation that meets the Group’s <strong>requirements and objectives</strong>.</p><h3><strong>Our process automation team</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/102/1*koGmmOUi3NCOn43tQAooQA.png" /><figcaption><a href="https://www.linkedin.com/feed/">Philipp Slupina</a><br> Lead​</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/102/1*O3g4PHnvUPMSEEZxnRvF_Q.png" /><figcaption><a href="https://www.linkedin.com/search/results/all/?fetchDeterministicClustersOnly=true&amp;heroEntityKey=urn%3Ali%3Afsd_profile%3AACoAAEacxHEBa-grIKjtNppimTurhwRMWq2PihA&amp;keywords=roland%20geropp&amp;origin=RICH_QUERY_TYPEAHEAD_HISTORY&amp;position=0&amp;searchId=affdb41c-73a4-4ca4-bf57-4239cf039ab7&amp;sid=CPY&amp;spellCorrectionEnabled=true">Roland Geropp</a><br> Automation Expert​</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/102/1*9fMwXO7gaDAd0cTATJGWgQ.png" /><figcaption><a href="https://www.linkedin.com/search/results/all/?fetchDeterministicClustersOnly=true&amp;heroEntityKey=urn%3Ali%3Afsd_profile%3AACoAAC3a-1sBmZfaNGaK5s7NgTAYeXXIZA0XSuU&amp;keywords=gregor%20holert&amp;origin=RICH_QUERY_TYPEAHEAD_HISTORY&amp;position=0&amp;searchId=67ee98d3-568a-4f75-8b21-9f18100429de&amp;sid=5Kb&amp;spellCorrectionEnabled=true">Gregor Holert</a><br> Automation Expert​</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/102/1*g1y7Q0UkK3p3iLEoiooOmA.png" /><figcaption>Aleksandar Culafic​<br> Working Student​</figcaption></figure><h3>Targeted audience</h3><p>This article targets anyone interested in implementing a greeting bot with using Power Automate to welcome new members to the company. In particular, the target group includes users who want to develop an automated solution to optimize the onboarding process and facilitate the integration of new team members. Especially for people who already have basic knowledge of Power Automate and have or want to develop concrete ideas for implementing a greeting bot.</p><h3>Requirements</h3><p><strong>Power Automate / Microsoft account: </strong>A Microsoft account is required to access Power Automate, with the appropriate permissions. This must be assigned.</p><p><strong>Microsoft Teams access: </strong>The bot will be used in a Microsoft Teams channel, so it must have the necessary access rights and permissions in Teams.</p><p><strong>Understanding of the workflows: </strong>A<strong> </strong>basic understanding of Power Automate and its functionality is important. This includes creating and managing flows and working with triggers and actions.</p><p><strong>Teams channel management: </strong>Basic knowledge of managing and configuring Teams channels is helpful.</p><p><strong>Testing the flow: </strong>Thoroughly test the flow in a test environment to ensure it works as desired and no errors occur.</p><p><strong>Documentation: </strong>Create documentation on how the bot works and how to configure the flow.</p><p><strong>Training:</strong> Train the relevant team members on how to use the bot and manage the flow.</p><p>Microsoft Teams serves as an example, but the bot can also be used in other communication tools as long as the corresponding Power Automate connectors are available.</p><h3>Implementation process</h3><p>The greeting bot is designed to be activated regularly, such as on the 1st or 15th of a month, to automatically introduce new team members in a Teams channel. First, the flow checks whether the current day is a working day to avoid sending messages on weekends or public holidays. If this condition is met, the flow accesses central data sources such as the Teams Admin Center, SharePoint, or other databases to retrieve the names, departments, and job titles of the new team members, for example.</p><p>Based on this data, the bot creates a structured and personalized welcome message that can be tailored to the company’s specific requirements. This message is then automatically posted in the Teams channel, informing the entire team about the new members. This fosters a consistent and friendly welcome culture and helps integrate new members into the team.</p><p>The development of this greeting bot begins with logging into Power Automate and creating a new scheduled flow. The flow is configured to trigger only on specified days and working days. Next, relevant data sources are integrated to obtain information about the new team members. Once the message template is created and the exact requirements have been agreed with the relevant department, the flow is tested in a test environment to ensure it works correctly. After successful testing, the flow is implemented in the production environment and monitored regularly to ensure smooth and efficient communication.</p><p>As visualized in <strong>Image 1</strong>, the graphic shows <strong>our implementation process for the greeting bot</strong>. It illustrates the individual steps and processes involved in setting up and implementing the bot. This serves as an example of our own implementation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*FqzAj6Dp83T25lRnq-Xcbw.png" /><figcaption>Image 1: Implementation process for the greeting bot</figcaption></figure><h3>Challenges</h3><p>Implementing such a project comes with many challenges, many of which only become apparent during execution.</p><p>One challenge is <strong>checking for weekends and public holidays</strong>, as shown in <strong>Image 2</strong>. The graphic shows a loop that stores the current date. This date is then compared with a list in SharePoint that contains public holidays. It also checks whether the saved date falls on a Saturday or Sunday.</p><p>If the date falls on a weekend or public holiday, a delay of 24 hours (one day) is inserted. Once the delay expires, the date is checked again. This process repeats until a working day is found. Only when a valid working day is identified does the flow continue.</p><p>This method ensures that the greeting bot sends messages only on working days, thus optimizing the effectiveness of the greetings.</p><p>Another challenge is <strong>formatting the post itself</strong>. The result is created in Power Automate using HTML, a programming language. Previous experience with HTML is a great advantage here. Variables such as name, date and lists must be formatted correctly and adapted accordingly in the programming language of Power Automate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/487/1*W7mwAg6ajTsWJaG5uW7rvQ.png" /><figcaption>Image 2: Review of working days and public holidays</figcaption></figure><p>This requires <strong>precise adjustment of the formatting</strong> in the text to ensure that the end result meets the desired standard. Incorrect formatting can lead to confusing or incomplete messages, so correct application is essential to achieve the result as shown in <strong>Image 3.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/849/1*XkwpvUOc5vhEmH_X7qclig.png" /><figcaption>Image 3: Sample greeting</figcaption></figure><h3>Conclusion</h3><p>The greeting bot developed with Power Automate is a valuable solution for improving the integration of new employees into teams. It automates the welcome process, creating consistently structured and personalized communication that ensures a positive welcome culture.</p><p>Automation reduces administrative burdens while ensuring that new team members are introduced efficiently and consistently. This contributes to a harmonious working environment and helps new colleagues integrate quickly into the team. The greeting bot therefore proves to be a useful tool for promoting internal communication and increasing efficiency within the company</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=91af1a5b5f02" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/by-aleksandar-culafic-91af1a5b5f02">By Aleksandar Culafic​</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Spotlight on Tech with SARAH]]></title>
            <link>https://medium.com/tech-p7s1/spotlight-on-tech-with-sarah-61ffc4b351a4?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/61ffc4b351a4</guid>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[women-in-tech]]></category>
            <category><![CDATA[ai-strategy]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Fri, 02 Aug 2024 10:48:10 GMT</pubDate>
            <atom:updated>2024-08-02T11:33:28.246Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Interested in some insights in our AI Strategy department? As Associate Director Data &amp; AI Strategy, Sarah has been part of ProSiebenSat.1 for six months. In the interview, she gives insights into why she appreciates coffee talks with our developers so much, what our “AI Ambassador Circle” is all about and what new colleagues can look forward to in her team.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XCUTKsljQ9op0S0yXa0RgA.png" /></figure><p><strong>Hi Sarah, you’ve been with us for half a year now. Can you tell us what you do at ProSiebenSat.1 and what you’ve experienced during your first time with us?</strong></p><p>I work in our AI Strategy Team as Associate Director Data &amp; AI Strategy, actually only for six months, although it feels much longer! I think that’s also because I joined a really great team and everyone took a lot of time for my onboarding.</p><p><strong>The topic of AI has only recently reached the “masses”. You have been personally interested in it for quite a while, but the fact that you are also involved in it professionally is also new to you. What did you do before, and what skills help you in your current role?</strong></p><p>Indeed, I have been fascinated by the topic of AI for quite some time. However, it has only become my focus professionally since joining ProSiebenSat.1. Previously, I was a founder of a Mental Health Tech startup, so I am used to diving into new topics, which I find very enjoyable. For example, during the early stages of the startup, we took over the production of short animated videos, and we taught ourselves everything we needed to know about it.</p><p>In my current role, entrepreneurial thinking is crucial, which is another skill I acquired during my time as a startup founder. I was used to identifying the most promising ideas from customer conversations and market analysis, quickly implementing them with the available resources and being agile in responding to market needs. And that is also a part of my current job.</p><p>Furthermore, my experiences as a research assistant at the Institute for Market-based Management at LMU and in Management Consulting have been helpful in the field of AI strategy. Statistics was my daily doing back then, and it also has a connection to AI since statistics plays a significant role in AI. That is something I find very exciting.</p><p><strong>Together with your team, you have launched the so-called “AI Ambassador Circle”. What is it exactly?</strong></p><p>When I started, the “AI Ambassador Circle” was actually my first big and exciting project. Its purpose is to coordinate AI activities, create synergies, and strengthen the AI competence of all employees. The “AI Ambassadors” were nominated by the leadership team of their respective departments. Currently, we have around 100 AI Ambassadors at ProSiebenSat.1 who exchange ideas with each other in our circle meetings to further advance AI topics within the company. Additionally, our “AI Ambassadors” receive regular AI training and bring that knowledge back to their teams.</p><p><strong>How else can we imagine working in the AI Strategy Team?</strong></p><p>We not only focus on so-called “enablement topics” like the “AI Ambassador Circle”, but also extensively on how we want to strategically position ourselves as a company to use AI in a targeted and responsible manner. This includes identifying concrete projects with a significant impact.</p><p>What’s special about our team is that we are part of the AI Hub and closely collaborate with product owners, developers, and data scientists from the AI Solutions team, who are responsible for a range of exciting topics. These include the “GenAI Playground”, which allows all P7S1 employees to use GenAI tools in a secure environment, generating AI solutions for content metadata, or innovative AI-based advertising formats like contextual targeting.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/570/0*XznB95gJLu2CuXS_" /><figcaption>Sarah, Associate Director Data &amp; AI Strategy</figcaption></figure><blockquote>“After my first half year here, I would say that I especially appreciate the culture and the people at P7S1. I really like how we interact with each other and the values we live as a team. My personal highlight is that I can easily go to the kitchen and have one of our developers, for example, briefly explain complex AI topics to me.” (laughs)</blockquote><p><strong>You are currently looking for an “Associate Manager AI Strategy (m/f/d)”. What should a person bring to be successful in this role?</strong></p><p>Yes, we are currently looking for an “<a href="https://jobs.prosiebensat1.com/job/M%C3%BCnchen-Associate-Manager-AI-Strategy-%28mwd%29-80802/798996002/">Associate Manager AI Strategy</a>” — a position that I find very exciting. The person will support us in the development and implementation of our AI strategy and also work on topics such as the “AI Ambassador Circle”. The most important criteria for this position, in our view, are a passion for AI and entertainment topics and a genuine desire to make a difference. Additionally, a very good university degree and professional experience in a strategy department or management consulting are important to us.</p><p><strong>What advice would you give to future colleagues?</strong></p><p>ProSiebenSat.1 has a lot of interesting people and characters. Therefore, I would advise everyone to participate in networking events and simply go for a coffee with people from other departments whom you haven’t met before. It has always been enriching for me.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*19sc2dy7sn2w0oAn1CfWQA.jpeg" /></figure><h4>📢Do the positions in Sarah’s team sound interesting to you? Then you will find all further information under the following links!</h4><p><a href="https://jobs.prosiebensat1.com/job/M%C3%BCnchen-Associate-Manager-AI-Strategy-%28mwd%29-80802/798996002/">Associate Manager AI Strategy (m/w/d)</a></p><p><a href="https://jobs.prosiebensat1.com/job/M%C3%BCnchen-Praktikum-AI-Strategy-%28mwd%29-80802/798945702/">Intern AI Strategy (m/w/d)</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=61ffc4b351a4" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/spotlight-on-tech-with-sarah-61ffc4b351a4">Spotlight on Tech with SARAH</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Kargo Series: Part 3 — Create Kubernetes Clusters at scale with Cluster API]]></title>
            <link>https://medium.com/tech-p7s1/kargo-series-part-3-create-kubernetes-clusters-at-scale-with-cluster-api-eff259d06874?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/eff259d06874</guid>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[cluster-api]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[platform-engineering]]></category>
            <category><![CDATA[kubernetes-cluster]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Wed, 24 Jul 2024 14:54:58 GMT</pubDate>
            <atom:updated>2025-01-16T16:26:57.773Z</atom:updated>
            <content:encoded><![CDATA[<h3>KARGO (pt 4) — Create Kubernetes Clusters at scale with Cluster API</h3><p>By <a href="https://www.linkedin.com/in/constantin-geisler-60bb35121/">Constantin Geisler</a></p><p>This third blog post is part of a series of posts on how we built KARGO — A Container focused Developer Platform at ProSiebenSat.1 Tech &amp; Services GmbH (the internal service provider at ProSiebenSat1). If you want to know more about <a href="https://medium.com/tech-p7s1/moving-up-the-stack-c680cebe234c">the decision to do so</a>, or <a href="https://medium.com/tech-p7s1/kargo-a-container-focused-developer-platform-on-aws-0bdc5262fa46">the overall architecture</a> take a look at our previous blog posts.</p><p>There are many ways to create an EKS cluster in AWS. Clicking around in the <a href="https://console.aws.amazon.com/eks/home#/clusters">AWS Management Console</a>, using the Amazon command line tools <a href="https://eksctl.io/">eksctl</a> or <a href="https://docs.aws.amazon.com/cli/latest/reference/eks/">awscli</a>. Certainly, using the de facto standard for infrastructure as code, terraform, was high on our list and would have been the way to go if there wasn’t something else that worked better for us.</p><p>We, the Platform Engineering &amp; Operations team, have built our own custom vanilla Kubernetes distribution, which creates a lot of work for the team just to maintain the platform so that all components are up-to-date. That was fine when we started this in 2016, but in 2024 that’s a solved problem. So the way to go is <a href="https://medium.com/tech-p7s1/moving-up-the-stack-c680cebe234c">moving up the stack</a>. We want to move away from maintaining a few large Kubernetes clusters to being able to deploy as many (smaller) clusters as we want. To achieve this, we need to automate everything. Creating or deleting a cluster needs to be easy and fast. It’s a challenging task to operate and maintain an unknown, but definitely growing number of Kubernetes clusters. We are lucky that there are tools in the <a href="https://www.cncf.io/">CNCF ecosystem</a> that make this possible.</p><h3>An Introduction to Cluster API</h3><p><a href="https://cluster-api.sigs.k8s.io/">Cluster API</a> (CAPI) is an open source project that provides declarative APIs and tooling to provision and manage Kubernetes clusters. It was originally created by <a href="https://github.com/kubernetes/community/tree/master/sig-cluster-lifecycle#readme">SIG Cluster Lifecycle</a> and has now become a subproject under <a href="https://github.com/kubernetes/community/blob/master/sig-architecture/README.md">SIG Architecture</a>. Cluster API makes it easier to manage Kubernetes clusters across different infrastructure providers like AWS, Azure, Google Cloud, vSphere etc. It’s the next level of Kubernetes cluster creation, configuration and management and is built on top of previous built cluster managers such as <a href="https://github.com/kubernetes/kops">kops</a> and <a href="http://kubicorn.io/">kubicorn</a>.</p><h3>Why use Cluster API for creating Kubernetes Clusters?</h3><p>We chose Cluster API to create EKS clusters because it uses Kubernetes style APIs, and we feel very comfortable with this approach. Since 2016, the team has been working with Kubernetes, so working with YAML files is something we are very used to. Also, the architecture of Cluster API, which is robust and also flexible, led to this decision.</p><h3>How the Cluster API works</h3><p>To get started with CAPI, the first thing you need is a Kubernetes cluster. Wait, what? I thought I would be creating my Kubernetes cluster with CAPI? Yes, but these are the so-called workload clusters. First, you need a management cluster.</p><p>Okay, so you need Kubernetes to deploy your Kubernetes. Well, okay. That sounds pretty much like the good old chicken-and-egg problem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TRcGYWwPh5h2VW01QJtwmw.jpeg" /><figcaption><em>The chicken egg problem (Picture by Constantin Geisler, via Bing — </em><a href="https://www.freepik.com/)"><em>https://www.freepik.com/)</em></a></figcaption></figure><p><strong>Let me explain our approach:</strong></p><p>We found a nice Linux distribution for hosting Kubernetes clusters called <a href="https://www.talos.dev/">Talos Linux</a>. We use Talos to create a management cluster on EC2 instances in AWS. The management cluster manages the lifecycle of the workload clusters and hosts the Cluster API controllers. It is the “brain”. A management cluster is also where one or more providers run and where resources such as machines are stored.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*k_GC6Gyc8r4MnMouAgRtEQ.png" /><figcaption><em>Figure 1 — Talos overview (Picture by the author, drawn with draw.io)</em></figcaption></figure><h3><strong>Architecture</strong></h3><p>The following picture shows the architecture of Cluster API:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*NjfxWyCskcINyUEtUJRAuw.png" /><figcaption><em>Cluster API concept (from </em><a href="https://cluster-api.sigs.k8s.io/user/concepts)">https://cluster-api.sigs.k8s.io/user/concepts)</a></figcaption></figure><p>At its core, the Cluster API uses <a href="https://kubernetes.io/docs/concepts/extend-kubernetes/api-extension/custom-resources/">custom resources</a> to define the desired state of a Kubernetes cluster. Here’s a breakdown of the key components:</p><ul><li>Cluster: The central resource that represents the entire cluster.</li><li>Machines: These are the worker nodes (or VMs) that run your applications.</li><li>MachineDeployments: These manage the lifecycle and scaling of a set of machines.</li><li>MachineSets: These ensure a stable set of replica Machines are running at any given time.</li></ul><p>With these resources, you can use familiar Kubernetes commands to manage your clusters, making the experience consistent and scalable.</p><p>To run workloads in AWS EKS it is necessary to install the AWS provider which is called <a href="https://github.com/kubernetes-sigs/cluster-api-provider-aws">CAPA</a>. With the running management cluster we are able to deploy as many workload clusters as we need.</p><h3>Cluster API in action</h3><p>To see how infrastructure resources are defined, let’s look at a simple example:</p><p>Defining a Cluster:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/923/1*1PYPItIU7dBADXc71Wthjw.png" /></figure><p>These examples highlight how you can streamline cluster management across different environments by using Kubernetes-native constructs to define and manipulate your clusters. We’ve created a helm chart, which is used to install all resources into separate AWS accounts with a single command. Before we can install a new workload cluster, we run Terraform to prepare the AWS account with all the necessary <a href="https://aws.amazon.com/iam/?nc1=h_ls">IAM permissions</a> for Cluster API and also for a set of basic tools (e.g. Cert Manager) that are installed on each workload cluster. To install the aforementioned basic tools, we also install <a href="https://fluxcd.io/">Flux CD</a> for our GitOps flow, so that all additional tools and applications are installed on each workload cluster without any further interaction from our side. There will be a separate blog post on Flux CD, stay tuned.</p><h3><strong>Pros of Cluster API</strong></h3><p>Let’s look at some of the benefits that the Cluster API brings to the table:</p><ul><li>Declarative syntax: Define your cluster in YAML and manage it with kubectl, just like you would with other Kubernetes resources.</li><li>Infrastructure agnostic: Cluster API works with public clouds, on-premises VMs, and even bare metal, providing a consistent API across different environments.</li><li>Scalability: Easily scale your clusters up or down as demand changes.</li><li>Version management: Simplify upgrades and rollbacks of cluster versions.</li><li>Consistency: Standardize cluster deployments, reducing the likelihood of environment-specific issues.</li><li>Community-driven: With the backing of the Kubernetes SIGs, the Cluster API is supported by a strong community.</li></ul><h3><strong>Cons of Cluster API</strong></h3><p>Despite its promise, the Cluster API isn’t without its challenges:</p><ul><li>Learning curve: New users may need to climb the learning curve of Cluster API concepts and resources.</li><li>Project maturity: Being relatively new, Cluster API may not be as mature as other Kubernetes tools and some features are still evolving.</li><li>Integration with existing tools: Not all existing tools and workflows are fully compatible with the Cluster API yet.</li></ul><h3>Conclusion</h3><p>Cluster API provides a powerful way to provision and manage Kubernetes clusters declaratively. With support for multiple infrastructure providers, it makes it easy to consistently deploy clusters across environments. The modular architecture allows for flexibility and extensibility. And overall, we feel like Cluster API is a very useful project for anyone who needs to operate Kubernetes at scale.</p><h3>More about KARGO</h3><ul><li><a href="https://medium.com/tech-p7s1/moving-up-the-stack-c680cebe234c">KARGO (pt 1) — Moving up the stack</a></li><li><a href="https://medium.com/tech-p7s1/kargo-a-container-focused-developer-platform-on-aws-0bdc5262fa46">KARGO (pt 2) — A container focused developer platform on AWS</a></li><li><a href="https://medium.com/tech-p7s1/exposing-workloads-on-eks-0ec39acd5fa9">KARGO (pt 3) — Exposing workloads on EKS</a></li><li><strong>KARGO (pt 4) — Create Kubernetes cluster at scale with ClusterAPI</strong></li><li><a href="https://medium.com/tech-p7s1/kargo-pt-5-managing-custom-eks-cluster-addons-with-flux-cd-0a9b6a706088">KARGO (pt 5) — Managing custom EKS cluster addons with Flux CD</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eff259d06874" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/kargo-series-part-3-create-kubernetes-clusters-at-scale-with-cluster-api-eff259d06874">Kargo Series: Part 3 — Create Kubernetes Clusters at scale with Cluster API</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Women in Tech x AI x Entertainment]]></title>
            <link>https://medium.com/tech-p7s1/women-in-tech-x-ai-x-entertainment-d9d55fde4012?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/d9d55fde4012</guid>
            <category><![CDATA[women-in-business]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[female-empowerment]]></category>
            <category><![CDATA[women-in-tech]]></category>
            <category><![CDATA[women-in-data]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Fri, 07 Jun 2024 13:20:35 GMT</pubDate>
            <atom:updated>2024-06-07T13:28:47.134Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pf5c69oVYumw592hw4f5PA.png" /></figure><p><strong>For the second time, we hosted the “Women in Big Data — P7S1 x AWS” afterwork event in our headquarters in Unterföhring this week.</strong></p><h4><strong>Women in Big Data x P7S1</strong></h4><p>The aim of the women’s network WiBD (Women in Big Data) is to enhance diversity within data and analytics, as well as across the broader spectrum of technology, by fostering leadership through technical education, career advancement, mentorship, and networking opportunities.</p><p>Ensuring a diverse work environment where women can succeed and grow is also of great importance to us at ProSiebenSat.1, alongside promoting the understanding that the tech industry is equally conducive to women’s success as it is to men’s.</p><h4><strong>On the agenda: Mingle, learn, contribute &amp; empower</strong></h4><p>We were very happy to welcome around 80 women from the tech industry at our campus in Unterföhring and bringing them together with some of our women in Tech from ProSiebenSat.1. The afterwork event commenced with a snack and mingling session, providing participants with the opportunity to acquaint themselves, engage in initial conversations, and of course capture moments of the evening at our designated photo corner.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M7DCigAP9C-_NEIQOSstmA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h2kzZ8kGskto527W1bpeEQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3H4NfntAjk2EDfi7yTBlNw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZGAGTnVtm0wTaf62ua1IAQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/604/1*95iBG-GEyyxDPhLpkdnqLQ.png" /></figure><h4><strong>🎬 Connecting AI with Entertainment: Our speakers shared their valuable insights</strong></h4><p>Hello from women’s network to women’s network. The evening startet with a warm welcome by Nahia Orduña (core member, Women in Big Data) and Stefanie Jenzsch (core member, F•Empowerment P7S1) followed by diverse talks and use cases about how AI is influencing and enabling the Media industry.</p><p><strong>The evening’s agenda</strong></p><p>👉 AI in Video Streaming — Role model talk with Marion von Nell, COO Joyn Gmbh</p><p>🙌 How can GenAI shape the Media and Entertainment Industry — Diana Cheng, Senior Solutions Architect AWS</p><p>👨‍🎓 AI Products in Media — Insights by Dr. Anca-Roxana Turdoran, Senior Data Scientist P7S1</p><p>✨ Diversity &amp; Inclusion in Media — Fireside chat by Anja Reinhard, SVP Diversity &amp; Inclusion and Jomy Jose, Chapter Director WiBD Munich</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X9xMEOv_5zTw3iuWxbswGw.png" /><figcaption>Marion von Nell, Chief Operating Offier Joyn GmbH</figcaption></figure><blockquote>“I was really impressed by the positive energy, creative ideas and questions from all participants. I am convinced that by working together and supporting each other, we can make great progress. Diversity is key to innovation and I am looking forward to seeing how each one of us will help shape and change the technology industry.”</blockquote><p>The final highlight on the agenda was a quiz followed by some drinks and music over the roofs, more interesting conversations and exchanges, and above all great women who all have something in common: their passion for data and technology, an open mindset and their support for each other.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/828/1*JFfgC1aNhLfWTfNYCOSanw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ii3o3c79SJyNF6ogIN8NOw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KySBkPbXR3ZDgm8mxSBNRg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fZvp9j9yYxsy4igl8GcwRQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/567/1*n1TCVRP13flJDE9hpraJjA.png" /></figure><p>A big thank you to all the amazing women for coming to our campus, sharing their great ideas and visions, and who made this event such a success!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d9d55fde4012" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/women-in-tech-x-ai-x-entertainment-d9d55fde4012">Women in Tech x AI x Entertainment</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The P7S1 OKR Day: Enhancing Collaboration and Innovation through OKRs ‍]]></title>
            <link>https://medium.com/tech-p7s1/the-p7s1-okr-day-enhancing-collaboration-and-innovation-through-okrs-eabcf4f28706?source=rss-b7364a9a56b1------2</link>
            <guid isPermaLink="false">https://medium.com/p/eabcf4f28706</guid>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[okr]]></category>
            <category><![CDATA[collaboration]]></category>
            <dc:creator><![CDATA[Tech@ProSiebenSat.1]]></dc:creator>
            <pubDate>Tue, 04 Jun 2024 10:53:14 GMT</pubDate>
            <atom:updated>2024-06-04T10:53:14.646Z</atom:updated>
            <content:encoded><![CDATA[<p>Recently, we hosted the second edition of our P7S1 OKR Day, bringing enthusiastic colleagues from various departments of the company together for a day filled with learning and collaboration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*cYdZjGwPtBNpxbBbgi6HAg.png" /></figure><p><strong>Oskar OKR: Our AI Assistant</strong></p><p>The day kicked off with an interesting presentation by Niklas (Associate ML Engineer) and Manuel (Data Scientist), who introduced our new OKR AI Assistant, called <em>Oskar OKR</em>. This AI tool provides teams with examples of potential Objectives and Key, based on the input they enter into to tool. The AI provides initial drafts, which can then be refined through further discussion and fine-tuning.</p><p><strong>Starting with OKRs: A Practical Guide</strong></p><p>Afterwards, Martin (Agile Coach) and Tobias (Agile Coach) presented a little “Hitch Hiker’s Guide” to get started with OKRs. They included essential information such as OKR Superpowers, Do’s and Don’ts, the OKR Cycle, and the principle of Top-Down/Bottom-Up Alignment. There was also time for questions and sharing experiences. To further support colleagues to have a successful start with OKRs, everyone could take home a printed version of the OKR One-Pager.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/701/1*zxMv-Tx4JA7zHY6mb7k9cw.png" /><figcaption>Martin and Tobias, Agile Coaches, presenting a concise ‘Hitch Hiker’s Guide’ for getting started with OKRs</figcaption></figure><p><strong>Outcome vs. Output</strong></p><p>Heiko (Agile Coach) then gave a short talk on the difference between Outcome and Output. Using vivid examples, he explained that it is worth thinking not only about an output (e.g. a list, a document or an app), but also about what it is used for and how it will change customer behavior (outcome). With the help of OKRs, it is therefore possible to not only focus on the working steps or To Do’s for a product, but also to formulate from the beginning, which value it has for the relevant target group.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/663/1*rycgX2IMjsL2JixfUj3utw.png" /><figcaption>Heiko (Agile Coach) highlights the difference between Outcome and Output</figcaption></figure><p>Another highlight of the day was the Open Space. This format allows any colleague, regardless of being an OKR expert or beginner, to bring up a topic for discussion. The goal is to engage as many colleagues as possible in meaningful exchanges, resulting in valuable discussions and great outcomes. To name some topics, we had sessions about “which topics are suitable for OKRs and which might be not?” or “How do we combine OKRs with other Frameworks?”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/496/1*xki2KkM1_oFgyYO4JpNMuw.png" /><figcaption>Open space session and a deep dive on the OKR AI Assistant</figcaption></figure><p>For Tobias, Agile Coach and co-organizer of the OKR Day, this year’s event was once again a success:</p><p><em>“With Objectives and Key Results, we are able to think iteratively even at a strategic level. We constantly ask for the ‘why’ rather than just for the ‘what’. This helps us maintain focus on the actual value we want to achieve with our products and projects, instead of being solely driven by our to-do lists. With the OKR Day, we always manage to connect colleagues from various areas of the company and spread this mindset throughout the entire company.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/525/1*iQO9CpPCkTRmRJT1XTOhVA.png" /><figcaption>Tobias, Agile Coach and co-organizer of the OKR Day</figcaption></figure><p>Many thanks to Martin, Jule, Tobias, Heiko, and Katja for organizing this inspiring event. 🙌</p><p>#P7S1 #OKR #Teamwork #Innovation #AI #OpenSpace #Learning #Collaboration</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eabcf4f28706" width="1" height="1" alt=""><hr><p><a href="https://medium.com/tech-p7s1/the-p7s1-okr-day-enhancing-collaboration-and-innovation-through-okrs-eabcf4f28706">The P7S1 OKR Day: Enhancing Collaboration and Innovation through OKRs 💡👩‍🎓</a> was originally published in <a href="https://medium.com/tech-p7s1">ProSiebenSat.1 Tech Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>