Welcome to Data Brew by Databricks with Denny and Brooke! In this series, we explore various topics in the data and AI community and interview subject matter experts in data engineering/data science. So join us with your morning brew in hand and get ready to dive deep into data + AI!
External link for Data Brew by Databricks
Data Brew by Databricks reposted this
Building a custom AI model doesn’t have to take massive labeled datasets or months of training. In this Data Brew episode, Predibase CTO and co-founder Travis Addair explains how reinforcement fine-tuning makes it possible to start fast, adapt seamlessly, and keep improving with human feedback — bringing advanced AI within reach for everyone. Watch the full episode ☕: https://lnkd.in/ghagMwa5
How can you prevent reward hacking in AI models—especially on complex tasks? In our latest #DataBrew episode, Travis Addair explains how curriculum learning—starting with simpler examples—and scoring methods like partial credit can help steer AI models toward correct solutions, where reward hacking is a risk. 🎧 Watch the full Data Brew episode: 🔹 Apple → https://lnkd.in/e5Emjy8F 🔹 Spotify → https://lnkd.in/eFzXHVjU 🔹 YouTube → https://lnkd.in/eZh63dMC Data Brew by Databricks is hosted by Brooke Wenig and Denny Lee. ☕ #databrew #databricks #reinforcementfinetuning
📣 New Data Brew Episode Out Now! In this episode, we sit down with Travis Addair, CTO and Co-Founder of Predibase, creators of the first reinforcement fine-tuning platform, to explore the future of specialized AI. Discover how reinforcement fine-tuning is revolutionizing model customization, enabling you to start fast, adapt to your unique data, and keep improving through human feedback. Learn how this breakthrough is making advanced AI accessible to everyone. 🎧 Watch the full Data Brew episode: 🔹 Apple → https://lnkd.in/e5Emjy8F 🔹 Spotify → https://lnkd.in/eFzXHVjU 🔹 YouTube → https://lnkd.in/eZh63dMC Data Brew by Databricks is hosted by Brooke Wenig and Denny Lee. ☕ #databrew #databricks #reinforcementfinetuning
Denny Lee caught up with Kilian Lieret Ph.D., Research Software Engineer at Princeton University and past Data Brew podcast guest, right after his packed 8 a.m. session at #DataAISummit! 🚀 Kilian discussed how SWE-bench and SWE-agent are pushing AI to tackle real-world software engineering—moving beyond toy problems to solve issues in massive codebases. SWE-agent’s flexibility makes it useful for both research and enterprise. He also teased what’s next: more training data, boosting small model performance, and new benchmarks like multimodal SWE-bench for front-end and image-based challenges. If you’re working on autonomous agents, check out these open challenges! Curious to learn more? Check out this Data Brew episode with Kilian Lieret Ph.D. & Carlos E. Jimenez 👉 https://lnkd.in/e9aEGKQp #databrew #swebench #sweagent #ai #databricks
Data Brew by Databricks reposted this
PM Director, Developer Relations at Databricks | Apache Spark™ & MLflow Contributor | Delta Lake & Unity Catalog Maintainer | Co-host of Data Brew Podcast | Advisor to Early-Stage AI & Data Startups
🌟 #DennysPick 🌟 Some people just raise the bar for an entire field, and for me, Sharon Zhou, PhD is one of them. I first met Sharon during the wild ride that was LLM Avalanche in 2023, and it’s been inspiring to watch her shape the GenAI community ever since. She’s not just a brilliant technologist; she’s an incredible educator and a true steward for everyone looking to learn and grow in AI. I even had the chance to sit down with Sharon for a Data Brew by Databricks podcast (watch here: https://lnkd.in/gscq5nUv), and every time I hear her break down the latest in fine-tuning, reliability, and performance, I walk away with new insights. If you want to learn from one of the very best, this is your chance. Here’s why Sharon’s session is on my must-see list: 🔹 Learn how to move beyond single-model solutions by composing compound AI systems with SLMs and highly accurate mini-agents. 🔹 Get the inside scoop on breakthrough techniques like memory RAG (for smarter, context-rich embeddings) and memory tuning with Mixture of Memory Experts (MoME) to reduce hallucinations and boost reliability. 🔹 The session is packed with real-world examples—text-to-SQL, factual reasoning, function calling, code analysis, and more—so you can see these methods in action across industries. 🔹 Most importantly, you’ll learn how to actually integrate mini-agents into agentic workflows for enterprise-ready, high-accuracy AI. Check out the full list of my picks ➡️ https://lnkd.in/gi9G25AF 📍 San Francisco 📅 June 9-12 🔗 Register today: https://lnkd.in/gcdz9iHC 🎟️ Want to join at half price? Comment “#DennysPick” and I’ll DM you the discount code! #DennysPick #AI #miniagents #aisystems #hallucinations #agenticai
Data Brew by Databricks reposted this
SWE-bench and SWE-agent are groundbreaking tools for evaluating and enhancing AI in software engineering. Princeton University Computer Science PhD Candidate Carlos E. Jimenez and Research Software Engineer Kilian Lieret Ph.D. share insights on: - AI tools in software engineering - Overcoming agent limitations, such as getting stuck in loops - The future of AI-powered code reviews Watch the full episode of #DataBrew: https://dbricks.co/3YpoYNm
Don’t miss this one at #DataAISummit (June 9-12)! 🎉 Kilian Lieret Ph.D. will share real insights on how AI can tackle real-world software engineering challenges. He’ll demo SWE-agent in action, discuss SWE-bench as a new benchmark for AI on large codebases, and give honest insights on today’s AI models. If you’re interested in agentic AI and practical tools for developers, be sure to catch this session!👇 Check out the details below. 🎥 For a sneak peek, watch our recent Data Brew episode with Kilian: https://lnkd.in/e9aEGKQp #databricks #sweagent #aimodels #AI #agenticai #databrew
PM Director, Developer Relations at Databricks | Apache Spark™ & MLflow Contributor | Delta Lake & Unity Catalog Maintainer | Co-host of Data Brew Podcast | Advisor to Early-Stage AI & Data Startups
🌟 #DennysPick 🌟 I recently had the chance to virtually sit down with Kilian Lieret Ph.D. from Princeton University on Data Brew by Databricks (check it out: https://lnkd.in/g_8nDkPx), and wow - he’s doing some seriously cool work! We geeked out over SWE-bench and SWE-agent, two groundbreaking tools that are pushing the boundaries of what AI can do in software engineering. Now, Kilian is taking the stage at Data+AI Summit to go even deeper, and I can’t recommend his session enough. Why am I excited about this #DataAISummit session? 🧩 Get an inside look at SWE-bench, the new benchmark for evaluating AI on real-world, large codebase tasks 🔍 Hear Kilian’s honest take on what today’s models can (and can’t) do—no sugarcoating 🛠️ See a hands-on demo of SWE-agent, and learn how to build and test your own software engineering agents 🚀 Walk away with practical insights into the challenges ahead and how you can start experimenting with agentic AI in your own workflows If you want to see where agentic AI is headed and how you can be part of it, add this session to your agenda! Check out the full list of my picks ➡️ https://lnkd.in/gi9G25AF 📍Join us in San Francisco, June 9-12! 🔗 Register today: https://lnkd.in/gcdz9iHC ✉️ Want to join DAIS at half price? Comment "#DennysPick” and I’ll DM you the discount code! #ai #data #agenticai #sweagent #softwareengineering
Data Brew by Databricks reposted this
Dipendra Misra and Alnur Ali explore the challenges of applying AI in enterprise environments and the tools that are helping to bridge the gap between cutting-edge research + real-world deployment. “One of the biggest challenges in making LLM-powered systems real is scale, both in terms of context size and high request volume.” 🎙️ A new episode of Data Brew by Databricks is out! Hosted by Brooke Wenig & Denny Lee. https://dbricks.co/42PgX5N
“Benchmarks should reflect the kinds of things customers actually care about.” In this clip, Pallavi Koppol shares why Databricks created DIBS—the Databricks Intelligence Benchmarking Suite—a domain-specific benchmark designed to evaluate models on real-world tasks like RAG, text-to-JSON, and function calling. 🎧 Watch the full Data Brew episode: 🔹 Apple → https://lnkd.in/eH5cjwQD 🔹 Spotify → https://lnkd.in/eywjAZ9m 🔹 YouTube → https://lnkd.in/e8sddV8W Data Brew by Databricks is hosted by Brooke Wenig and Denny Lee. #databrew #databricks #benchmarking #DIBS
📣 NEW #DataBrew Episode: Benchmarking LLMs for Enterprise AI In this episode, Databricks Research Scientist Pallavi Koppol dives into the critical role of domain-specific intelligence in large language models. Discover why enterprise-grade AI demands more than general benchmarks—and how to evaluate models where it truly counts. Highlights include: 🌟 Why benchmarking LLMs for domain-specific tasks is critical for enterprise AI 🌟 An introduction to the Databricks Intelligence Benchmarking Suite (DIBS) 🌟 Evaluating models on real-world applications like RAG, text-to-JSON, and function calling 🌟 The evolving landscape of open-source vs. closed-source LLMs 🌟 How industry and academia can collaborate to improve AI benchmarking 🎧 Listen on your favorite platform: 🔹 Apple → https://lnkd.in/eH5cjwQD 🔹 Spotify → https://lnkd.in/eywjAZ9m 🔹 YouTube → https://lnkd.in/e8sddV8W Data Brew by Databricks is hosted by Brooke Wenig and Denny Lee. #databricks #llm #enterpriseai