--- title: "Git Smart: AI Agents in Data Engineering" newsletter: "MLOps Community" date: 2025-05-01 source: https://aaif.live/newsletters/mlopscommunity/2025-05-01-git-smart-ai-agents-in-data-engineering --- # Git Smart: AI Agents in Data Engineering *Plus, how graphs and LLMs improve data analytics, an agentic workflow for insights from Google Chat, Confessions, and Hidden Gems.* *MLOps Community — Agentic AI Foundation, 2025-05-01* And I thought I'd made a friend [https://go.mlops.community/89fdw4]. ## AI Data Engineers - Data Engineering After AI Stable wifi for a Zoom call. A git merge with no conflicts. Fixing a bug and nothing else breaks. We've all had those “no way this just worked” moments. Vikram shared a new one when their agent pulled off a full Airflow-to-Databricks pipeline using Spark. The system connects to tools like Airflow, GitHub, and Postgres, and safely tests changes in staging. Instead of overloading the context window, it pulls only what’s needed. * Agents plan before acting: Users review and edit steps before execution * Custom benchmarks: Eval runs track real outcomes like schema alignment Click below to listen – no surprises when this one just works. Video [https://go.mlops.community/8p7h8z] || Spotify [https://go.mlops.community/r5jzvl] || Apple [https://go.mlops.community/te0ifg] [https://go.mlops.community/te0ifg](https://go.mlops.community/te0ifg) [https://go.mlops.community/8p7h8z](https://go.mlops.community/8p7h8z) [https://go.mlops.community/r5jzvl](https://go.mlops.community/r5jzvl) ## GraphBI: Expanding Analytics to All Data Through the Combination of GenAI, Graph, & Visual Analytics The table was a mess because of the leak in the pipeline, and everyone had to wait because the server was overloaded. It could be a bad day in the office, or a bad night at a restaurant. Either way, it shows the core challenge with multi-domain data: meaning depends entirely on context. This chat was great for showing how well graphs pair with LLMs, especially when working with messy, sensitive information. We covered tokenising PII for secure processing, graph-based approaches to financial crime, and how graphs support more intuitive, context-rich reasoning. Graph structure becomes especially useful when reshaping unstructured inputs: * Visual graph mutation helps distill complex data into domain-specific insights * Collapsing noisy nodes into abstracted edges makes graphs easier to work with, without losing nuance Click below to listen for the context why we talked about towels at Brazilian parties. Video [https://go.mlops.community/7gjpku] || Spotify [https://go.mlops.community/ktrjex] || Apple [https://go.mlops.community/ytoxzy] [https://go.mlops.community/ytoxzy](https://go.mlops.community/ytoxzy) [https://go.mlops.community/7gjpku](https://go.mlops.community/7gjpku) [https://go.mlops.community/ktrjex](https://go.mlops.community/ktrjex) ## May 28 WORLD TOUR OF AI AGENTS We’re launching a global series of meetups, starting at the legendary Hibernia in San Francisco on May 28 with the AI Agent Builders Summit -a jam-packed kickoff for the agents era. Expect 30+ booths showing off agents actually in production, plus stories, demos, and ideas from builders who’ve been deep in the trenches. Want a booth? If you’re building something clever, chaotic, or just straight-up useful, we want to see it. This is your moment to show off what your agent can do — and meet other builders doing the same. Thinking about sponsoring? We’re teaming up with folks who want to support the agent frontier and be part of this wild ride. If that sounds like you, hit us up. More dates and cities are on the way, and word on the street is we’ve got our eyes on some beautiful venues! Tell us if you want in - we’re curating the most electric show-and-tell the AI world’s ever seen. Register here [https://go.mlops.community/WorldTourSF] [Register here](https://go.mlops.community/WorldTourSF) ## Job of the Week [https://go.mlops.community/47zonj](https://go.mlops.community/47zonj) [https://go.mlops.community/MLConfess](https://go.mlops.community/MLConfess) ## BKFC: An Agentic Workflow for Gathering Knowledge from Google Chat Hidden between the 57th "Sounds good to me" and the 89th "Can you resend the link?", that important update got buried. This blog looks at how to dig it back up. BKFC is a Python notebook that extracts structured insights from Google Chat using a simple agentic workflow: Fetch, Process, Analyze, Report. It retrieves recent conversations, groups messages by space, and prepares clean text blocks for analysis. Vertex AI’s Gemini model processes the text using a Pydantic schema to pull out summaries, project updates, technical insights, and action items. A key part of BKFC’s approach is how it prepares the data before analysis: * Chronological grouping: Messages are sorted by thread and space to maintain context. * Structured extraction: Outputs follow a consistent schema, improving reliability for downstream use. Click below to get your chat back on track. Read it here [https://go.mlops.community/821yh5] [Read it here](https://go.mlops.community/821yh5) ## Hidden Gems ## HERE TO HELP Before you go, here are three ways I can help - just hit reply: * Curated intros to other community members * What problems are you dealing with? Let me help you find the best solutions through my network * Looking to augment your staff for an MLOps or AI project? I got you covered Thanks for reading, catch you next time! --- Source: https://aaif.live/newsletters/mlopscommunity/2025-05-01-git-smart-ai-agents-in-data-engineering