Computer Science

Our next meet-up is scheduled for March 18th....and we promise it will be a blast! ​Join us for a no-fluff, practitioner-led deep dive into two critical LLM challenges with focus on quantization and cheat-proof evaluations . ​This time we will handle a pannel discussion with two experts in the area - David Kоchoanov and Yordan Darakchiev. ​Pannel 1: Smarter AI at lower cost: How quantization makes LLMs more efficient? ​What can you expect? ​LLMs guzzle RAM and GPU like crazy. Quick rundown: inference modes, bottlenecks, the math behind it. ​Hands-on: when to use each and how to benchmark properly. ​The talk will provide a quick review of LLM inference modes, current bottlenecks, and numerics. The main goal is to explain post-training quantization methods that are most commonly supported by modern software and hardware, and give practical recommendations for their use and model benchmarking. We'll also provide a brief summary of the academic literature and discuss quantization-aware training and distillation. ​Who will present? ​David Kochanov has over 10 years of experience in computer vision research and development. In his current role at CARIAD (Volkswagen Group), he works on autonomous driving and driver assistance systems, with a focus on model deployment for real-time inference, including quantization and architectural optimizations. ​Pannel 2: Everybody Lies: How Language Models (and People) Cheat ​What can you expect? ​Teaching computers and people have the same temptations - students game grades, LLMs game benchmarks. We grade LLMs like exams: bad criteria reward test-taking tricks, not truth. The talk will present specially built for the pannel small BI-style benchmark (SQL, data summaries, and insight generation) with multiple LLM "judges" grade multiple models. With only couple of stress tests on the judges with small changes to the grading scheme all were broken - creating pure chaos! ​Rankings flip depending on who grades, response length beats correctness, and judges favour familiar model "accents". ​We'll end with an "academic integrity" toolkit for LLM evaluation: ensembles, invariance checks, length controls, and lightweight human calibration to make our LLM metrics harder to cheat on, and easier to trust. ​Who will present? ​Yordan Darakchiev is a data scientst and technical trainer. He has helped various teams turn data and ML into modern, measureable, production systems. His specialty are vision and language models, with a sharp focus on evaluation - fairness, privacy, trustworthiness, and robustness. He also teaches math and machine learning and has helped kickstart the career of more than 1000 data and ML professionals. ​Our next meet-up is scheduled for March 18th....and we promise it will be a blast! ​Join us for a no-fluff, practitioner-led deep dive into two critical LLM challenges with focus on quantization and cheat-proof evaluations . ​This time we will handle a pannel discussion with two experts in the area - David Kоchoanov and Yordan Darakchiev. ​Pannel 1: Smarter AI at lower cost: How quantization makes LLMs more efficient? ​What can you expect? ​LLMs guzzle RAM and GPU like crazy. Quick rundown: inference modes, bottlenecks, the math behind it. ​Hands-on: when to use each and how to benchmark properly. ​The talk will provide a quick review of LLM inference modes, current bottlenecks, and numerics. The main goal is to explain post-training quantization methods that are most commonly supported by modern software and hardware, and give practical recommendations for their use and model benchmarking. We'll also provide a brief summary of the academic literature and discuss quantization-aware training and distillation. ​Who will present? ​David Kochanov has over 10 years of experience in computer vision research and development. In his current role at CARIAD (Volkswagen Group), he works on autonomous driving and driver assistance systems, with a focus on model deployment for real-time inference, including quantization and architectural optimizations. ​Pannel 2: Everybody Lies: How Language Models (and People) Cheat ​What can you expect? ​Teaching computers and people have the same temptations - students game grades, LLMs game benchmarks. We grade LLMs like exams: bad criteria reward test-taking tricks, not truth. The talk will present specially built for the pannel small BI-style benchmark (SQL, data summaries, and insight generation) with multiple LLM "judges" grade multiple models. With only couple of stress tests on the judges with small changes to the grading scheme all were broken - creating pure chaos! ​Rankings flip depending on who grades, response length beats correctness, and judges favour familiar model "accents". ​We'll end with an "academic integrity" toolkit for LLM evaluation: ensembles, invariance checks, length controls, and lightweight human calibration to make our LLM metrics harder to cheat on, and easier to trust. ​Who will present? ​Yordan Darakchiev is a data scientst and technical trainer. He has helped various teams turn data and ML into modern, measureable, production systems. His specialty are vision and language models, with a sharp focus on evaluation - fairness, privacy, trustworthiness, and robustness. He also teaches math and machine learning and has helped kickstart the career of more than 1000 data and ML professionals. Agenda: 18:30-19:00 - Welcome drinks 19:00-20:00 - Pannel 1 & Pannel 2 20:00-20:30 - Q&A 20:30-21:30 - Networking ​Note: Event will be handled in English. ​Event is free but registration is mandatory: https://luma.com/8bbbxx1t

Our next meet-up is scheduled for March 18th....and we promise it will be a blast! ​Join us for a no-fluff, practitioner-led deep dive into two critical LLM challenges with focus on quantization and cheat-proof evaluations . ​This time we will handle a pannel discussion with two experts in the area - David Kоchoanov and Yordan Darakchiev. ​Pannel 1: Smarter AI at lower cost: How quantization makes LLMs more efficient? ​What can you expect? ​LLMs guzzle RAM and GPU like crazy. Quick rundown: inference modes, bottlenecks, the math behind it. ​Hands-on: when to use each and how to benchmark properly. ​The talk will provide a quick review of LLM inference modes, current bottlenecks, and numerics. The main goal is to explain post-training quantization methods that are most commonly supported by modern software and hardware, and give practical recommendations for their use and model benchmarking. We'll also provide a brief summary of the academic literature and discuss quantization-aware training and distillation. ​Who will present? ​David Kochanov has over 10 years of experience in computer vision research and development. In his current role at CARIAD (Volkswagen Group), he works on autonomous driving and driver assistance systems, with a focus on model deployment for real-time inference, including quantization and architectural optimizations. ​Pannel 2: Everybody Lies: How Language Models (and People) Cheat ​What can you expect? ​Teaching computers and people have the same temptations - students game grades, LLMs game benchmarks. We grade LLMs like exams: bad criteria reward test-taking tricks, not truth. The talk will present specially built for the pannel small BI-style benchmark (SQL, data summaries, and insight generation) with multiple LLM "judges" grade multiple models. With only couple of stress tests on the judges with small changes to the grading scheme all were broken - creating pure chaos! ​Rankings flip depending on who grades, response length beats correctness, and judges favour familiar model "accents". ​We'll end with an "academic integrity" toolkit for LLM evaluation: ensembles, invariance checks, length controls, and lightweight human calibration to make our LLM metrics harder to cheat on, and easier to trust. ​Who will present? ​Yordan Darakchiev is a data scientst and technical trainer. He has helped various teams turn data and ML into modern, measureable, production systems. His specialty are vision and language models, with a sharp focus on evaluation - fairness, privacy, trustworthiness, and robustness. He also teaches math and machine learning and has helped kickstart the career of more than 1000 data and ML professionals. ​Our next meet-up is scheduled for March 18th....and we promise it will be a blast! ​Join us for a no-fluff, practitioner-led deep dive into two critical LLM challenges with focus on quantization and cheat-proof evaluations . ​This time we will handle a pannel discussion with two experts in the area - David Kоchoanov and Yordan Darakchiev. ​Pannel 1: Smarter AI at lower cost: How quantization makes LLMs more efficient? ​What can you expect? ​LLMs guzzle RAM and GPU like crazy. Quick rundown: inference modes, bottlenecks, the math behind it. ​Hands-on: when to use each and how to benchmark properly. ​The talk will provide a quick review of LLM inference modes, current bottlenecks, and numerics. The main goal is to explain post-training quantization methods that are most commonly supported by modern software and hardware, and give practical recommendations for their use and model benchmarking. We'll also provide a brief summary of the academic literature and discuss quantization-aware training and distillation. ​Who will present? ​David Kochanov has over 10 years of experience in computer vision research and development. In his current role at CARIAD (Volkswagen Group), he works on autonomous driving and driver assistance systems, with a focus on model deployment for real-time inference, including quantization and architectural optimizations. ​Pannel 2: Everybody Lies: How Language Models (and People) Cheat ​What can you expect? ​Teaching computers and people have the same temptations - students game grades, LLMs game benchmarks. We grade LLMs like exams: bad criteria reward test-taking tricks, not truth. The talk will present specially built for the pannel small BI-style benchmark (SQL, data summaries, and insight generation) with multiple LLM "judges" grade multiple models. With only couple of stress tests on the judges with small changes to the grading scheme all were broken - creating pure chaos! ​Rankings flip depending on who grades, response length beats correctness, and judges favour familiar model "accents". ​We'll end with an "academic integrity" toolkit for LLM evaluation: ensembles, invariance checks, length controls, and lightweight human calibration to make our LLM metrics harder to cheat on, and easier to trust. ​Who will present? ​Yordan Darakchiev is a data scientst and technical trainer. He has helped various teams turn data and ML into modern, measureable, production systems. His specialty are vision and language models, with a sharp focus on evaluation - fairness, privacy, trustworthiness, and robustness. He also teaches math and machine learning and has helped kickstart the career of more than 1000 data and ML professionals. Agenda: 18:30-19:00 - Welcome drinks 19:00-20:00 - Pannel 1 & Pannel 2 20:00-20:30 - Q&A 20:30-21:30 - Networking ​Note: Event will be handled in English. ​Event is free but registration is mandatory: https://luma.com/8bbbxx1t

We are hosting our next meetup in Sofia! **Address:** Barter Community Hub \| Puzl\, 47 Cherni Vrah BLVD\, Sofia\, Bulgaria https://www.barter-hub.com **Agenda:** 17.45: Doors open, join us for snacks and drinks 18:15: Building Data Products with Logstash 19.00: How Elastic Streams Simplifies Log Data Management 19.45: Networking 20.00: Close **Talks:** **Building Data Products with Logstash** This talk will demonstrate how Logstash can be used to pull data, perform cleanup and enrichment, and create a data product (e.g., asset tracking) visualized in Kibana. *Speaker: Tsvetan Paskov, Full-stack Developer, DevOps, Founder @Move2Chat* **How Elastic Streams Simplifies Log Data Management** Learn how to manage log data more efficiently using Elastic Streams. This talk covers the challenges of unstructured logs, the benefits of schema-on-ingest, and new stream processing capabilities for transforming and structuring log messages. You’ll see how to use AI-assisted Grok pattern suggestions, conditional processing, and the schema editor to map fields accurately. We also explore how to monitor data quality through degraded and failed document tracking, and how to set custom retention policies. Finally, you’ll get an overview of Wired Streams, Elastic’s unified log ingestion model that simplifies routing and stream partitioning post-ingestion. *Speakers: Boris Kirov (Observability - Senior Product Designer) and Giorgos Bamparopoulos (Senior Manager Software Engineering) @Elastic*

📌 Let's meet in **Sofia** on **March 19th 2026** for an interactive conference and an enriching networking session! Come and discover Odoo, the all-in-one integrated software that will simplify your day-to-day business. Say goodbye to double coding, lost invoices and budgets; create your website in a few clicks... Odoo offers you a wide range of tools built for IT.🧐 🔎 **AGENDA** 🔍 18h00: Welcome! 18h15: Conference - Interactive demo & tips to digitize your company. 20h00: Networking - Snacks and drinks. ➡️ Are you convinced? You can subscribe at the following link: https://www.odoo.com/r/RfP The event is free, but registration is mandatory.❗ **Important:** 📍 Where? Sofia Grand Hotel - 1, General Gurko St str, 1000 Sofia, Bulgaria 📅 Date? Thursday, March 19th , 2026 📝 Registration? https://www.odoo.com/r/RfP 💰 Price? FREE! Hope to see you soon!

**Founders Running Club** (FRC) brings founders, investors, tech, creative people and startup enthusiasts together for weekly easy runs and networking. We like to be comfortable when we run and finish with coffee and conversations. Choose your pace or follow a pacer—pets, friends, family, are welcome. 🗓️ Launched in San Francisco, July 16, 2022 🌍 Now in 35+ cities 📅 Running + Networking events + Community **Join the community** [http://foundersrc.com/chats](http://foundersrc.com/chats) **Stay updated**: Instagram [http://instagram.com/foundersrc/](http://instagram.com/foundersrc/) Podcast [http://podcast.foundersrc.com/](http://podcast.foundersrc.com/) LinkedIn [http://linkedin.com/company/foundersrc/](http://linkedin.com/company/foundersrc/) Strava [http://strava.com/clubs/foundersRC](http://strava.com/clubs/foundersRC) Website [http://foundersrc.com/](http://foundersrc.com/)

Discover Kiro, AWS’s innovative development environment that adapts to your workflow. Whether you prefer specification-driven development or a more intuitive, vibe-based approach, Kiro empowers you to build applications your way. In this session, speakers will explore Kiro’s core functionality and its evolution through re:Invent 2025, along with demonstrations of prompts used to create APIs, hooks, and steering documents. In the latter part of the session, watch as we build a complete example from scratch, showcasing Kiro’s agentic AI workflow in action for spec-driven development and vibe coding. What You'll Learn - Core Kiro concepts and development philosophy Feature evolution: pre and post re:Invent 2025 Practical examples: API creation, hooks, and steering docs Live coding demonstration with real-time problem solving Who Should Attend - Developers, solutions architects, and technical leaders interested in modern AWS development tools and AI-assisted coding workflows. **Speakers Bio:** Matthew Jorat: Matthew Jorat is a Customer Solutions Manager at AWS with nearly three decades of expertise in AI, cloud transformation, migration, modernization, IT, technology, and business. His career spans infrastructure engineer, solutions architect, security auditor, customer success director, support and services director, and operations officer. He was an early adopter of AWS services as a customer, and his current role at AWS uniquely blends his technical and business skills as he partners with executive teams to translate strategic priorities into measurable business outcomes through cloud adoption, AI implementation, and operational excellence. Matt leads end-to-end AI initiatives for enterprise customers, identifying high-impact use cases and maintaining relentless focus on measurable ROI. He holds eight AWS certifications and serves as Adjunct Professor at Franklin University teaching cloud computing and AWS architecture. His expertise encompasses GenAI, agentic AI systems, migration, modernization, and optimization strategies that drive transformation at scale. [https://www.linkedin.com/in/mattjorat/](https://www.linkedin.com/in/mattjorat/) Girish Bhatia: Girish Bhatia is a seasoned technology delivery leader with deep expertise in cross-domain program management. He holds multiple AWS certifications, including AWS Certified Solutions Architect – Associate, AWS Certified Developer – Associate, and AWS Generative AI Practitioner, and is also an AWS Community Builder. Girish brings over two decades of experience in the information technology industry. Throughout his career, Girish has led major technology-driven business transformation from modernizing on-premises applications to driving large-scale cloud migrations. His work includes API-based integration platforms, data-center and mainframe modernization, and enterprise cloud adoption initiatives. Based in Columbus, Ohio, Girish works at a tech-forward fintech company. He is passionate about AWS serverless architectures, Generative AI and emerging cloud technologies. An avid learner and builder, he continues to explore and apply the latest innovations in AWS to help organizations accelerate their digital transformation. https://www.linkedin.com/in/girishbhatia/[https://www.linkedin.com/in/girishbhatia/](https://www.linkedin.com/in/girishbhatia/) **THANK YOU** *Franklin University* for hosting our meetup! To learn more about *Franklin University*, please visit their website: https://www.franklin.edu/ **DIRECTIONS** Franklin University Fisher Hall 300 E. Main St, Columbus, OH 43215 Map: https://maps.app.goo.gl/jxjBA2hUmS5qrvhq8 Parking is FREE! Please park in Lot C in front of Fisher Hall. See attached map. NOTE: Map the address only. When mapping with Google Maps it may use the Fisher Hall at OSU, which is NOT correct. **Want to sponsor the pizza and/or bar tab?** Please contact me if you would like to sponsor this meetup's pizza and/or bar tab: angelo@mandato.com

**Important time note:** Please plan on arriving between 5:30 and 6:00 as the elevators lock after 6 and you'll need to message us and we'll need to come get you. The building address is 4450 Bridge Park The entrance is 6620 Mooney St, Suite 400 **Abstract** Web Forms gave .NET developers a powerful abstraction for building reusable UI controls long before design systems were a thing. Web Components finally bring that same idea to the browser natively. This session explores how the Web Forms mindset translates into Custom Elements, Shadow DOM, and HTML templates. We’ll examine how Web Components enable design systems that work across frameworks, how they differ from server‑driven controls, and why they’re becoming a foundational layer for modern UI. Whether you’re maintaining legacy apps or building greenfield projects, you’ll leave with a practical understanding of how to apply familiar patterns in a modern, standards‑based way. **YouTube Link** TBA

This meeting will be a discussion of the book "the Science of Aliens" by Mark Brake

OpenClaw, NemoClaw, personal assistants, 24/7 agents. What does it all mean? How does it all work and what should we know about it? Whether you’re deep in the field or just getting curious, come connect with others building and exploring AI in Columbus. Sponsored by [Transform Labs](https://www.transformlabs.com/services)

Friday Night French Conversation Hour Bonjour tout le monde! (English Below) Je vous propose une soirée, ce vendredi qui vient, à l'hôtel Junto à Columbus. 19hr. 20 Mars (Vendredi) 2026. C'est très bien située avec de l'espace extérieur et intérieur, un feu, un café (cafe se ferme à 19h; arrivez tot si ca vous interesse), un resto, et un bar. Le parking est payé est juste en face au musée des sciences COSI. Il y'en a aussi dans les rues mais un peu plus loin. Vous n'êtes pas obligé d'acheter quelque chose, car l'ensemble de l'espace c'est vraiment un grand lobby et de l'hôtel, mais c'est cool. J'ai parlé avec le management et ils sont ravis de nous accueillir, mais il n'y a pas de place particulière réservé pour nous. Ce sera à partir de 19h jusqu'à ? Moi, je suis Brandon. J'aimerais que ce soit un évènement mensuel et si l'espace ne vous convient pas, on peut discuter d'autres options. Je vais porter une casquette bleue et blanche qui dit "Québec." Venez nombreux! A tres bientot \-\-\-\- Hello everyone! It’s been a long time since we’ve had a Friday night French meetup. I suggest that we meet at the Junto Hotel in Columbus this coming Friday, 20 March (friday) 2026 at 7PM. There’s a great spot with both outdoor and indoor areas, a café (which closes at 7 p.m., come a bit early if that interests you), a restaurant, and a bar. Parking is paid and is located right across from the science museum called COSI. There is additional free and paid street parking around at varying distances. It’s a bit expensive, but you don’t have to buy anything since the area is really just a large hotel lobby. It's nice though. I spoke with management, and they’re happy to welcome us, though there’s no specific space reserved just for us. It’ll start at 7 p.m. and go until whenever. I’d like this to become a monthly event, and if this space doesn’t suit us well, we can discuss other options. Speaking and understanding French well is ideal but there may be learners there as well. I’m Brandon. I’ll be wearing a blue and white cap that says “Québec.” Come find us.

**Improving Office in Franklinton** Physical location: Improving Office 330 Rush Alley Suite #150 Columbus, OH 43215 Schedule: 6:00 p.m.: Socialize, eat, and drink. Improving will be providing pizza and beverages. 6:30 to 8:00 pm. Main meeting and presentation(s). Topic: Type Annotations with John Cassidy We meet on the last Monday of each Month. Presentations are given by members and friends of this group. If you would like to do a presentation (small or large) on a python topic, please contact Central OH Python at centralohpython@gmail.com

Want to be a speaker? submit your talk to our Call for Presenters!!! https://sessionize.com/cbus-hug-2026/