--- title: "Agents" newsletter: "MLOps Community" date: 2025-05-15 source: https://aaif.live/newsletters/mlopscommunity/2025-05-15-agents --- # Agents *Plus - MLOps with Databricks, Iceberg, MCP, and MLOps: Bridging the gaps for Enterprise, GenV: An Agentic Workflow for Actionable Insights from Google Meet Recordings, and Blueprints for the Agentic Era: Inside the Revolution Transforming Enterprise AI.* *MLOps Community — Agentic AI Foundation, 2025-05-15* https://www.openxdata.ai/?utm_source=mlopscommunity&utm_medium=email&utm_campaign=2025_05_openxdata&utm_content=20250515_newsletter ## AI, Marketing, and Human Decision Making Do I like it when people ask and then answer their own questions? No, I don't. But not as much as Fausto hates it as a dead giveaway for Chat-generated content. We got into that while talking about how GenAI is reshaping marketing workflows – not just for manual asset creation, but as part of automated pipelines that swap out creatives and run real-time A/B tests across platforms. That raised bigger questions around how GenAI can turn vague ideas into working products faster, especially when paired with tools like Cursor, Whisper variants, and MCPs. One area that’s really taken off is UI generation: * Tools like 21stDev and browser-based MCPs make it easy to go from vague visual ideas to functioning components and app structures. * Scraper MCPs can pull in fresh SDK docs and turn them into structured references or rules inside your dev environment. We also explored how decision-making is shifting – from users to agents – and what that means for marketing, creativity, and even social recognition in human workflows. Is it difficult to hear the full episode? No, just click below to listen! Video [https://go.mlops.community/4kvfhj] || Spotify [https://go.mlops.community/oj9dl4] || Apple [https://go.mlops.community/uf0iq6] [https://go.mlops.community/uf0iq6](https://go.mlops.community/uf0iq6) [https://go.mlops.community/4kvfhj](https://go.mlops.community/4kvfhj) [https://go.mlops.community/oj9dl4](https://go.mlops.community/oj9dl4) ## MLOps with Databricks Home Depot, McDonald’s, Target, Databricks - consistent, familiar, and built to scale. It might not be the most flexible, but Databricks has become the go-to ML platform in many orgs because it fits neatly into existing workflows and clears procurement hurdles. Maria shared where it works well, and where it still gets in the way - especially the notebook-heavy development process. One area she recommends more teams use is Asset Bundles: * They package workflows, code, and configs in one place, making deployments far smoother. * They also help during development, letting you run code locally in a reproducible way before pushing to Databricks. The Feature Store shows promise, but it’s still rough around the edges. Versioning is limited, runtime behavior can be inconsistent, and it’s less friendly for non-Spark setups. Grab a Starbucks and click below to listen. Video [https://go.mlops.community/sjt2jg] || Spotify [https://go.mlops.community/3djb35] || Apple [https://go.mlops.community/us9wod] [https://go.mlops.community/us9wod](https://go.mlops.community/us9wod) [https://go.mlops.community/sjt2jg](https://go.mlops.community/sjt2jg) [https://go.mlops.community/3djb35](https://go.mlops.community/3djb35) ## Iceberg, MCP, and MLOps: Bridging the gaps for Enterprise ## MINI SUMMIT Like organizing a night out with more than two people, MLOps setups can be complex and fragile – but this session looked at how enterprise teams are trying to untangle that. Caleb focused on integrating ML workflows directly into the data platform to improve governance and scale. Hamza argued for flexible orchestration over enforced tooling, and Simba showed how structured features can be surfaced to LLM agents using MCP. The Snowflake example offered a clear look at how some teams are managing dev–prod separation: * Sandbox schemas with clear CI/CD paths let teams iterate safely before promoting to production. * Role-based access controls are applied to ML objects like features and models for consistent governance. Click below to watch and simplify things. For your MLOps setup, not your friends. Watch here [https://go.mlops.community/yasg2x] ## Job of the Week [https://go.mlops.community/ytgb7u](https://go.mlops.community/ytgb7u) [https://go.mlops.community/MLConfess](https://go.mlops.community/MLConfess) ## Blueprints for the Agentic Era: Inside the Revolution Transforming Enterprise AI To get you in the mood for our World Tour, here’s a recap from our recent AI Agents Masterclass in San Francisco. It breaks down how agents are already in production across industries, with companies moving quickly from pilots to scaled systems. CrewAI, LlamaIndex, and Lambda focused on orchestration, observability, and building infrastructure that avoids vendor lock-in. One standout session explored how to make multi-agent systems actually work: * Lambda used crews of specialized agents to automate BI reports, cutting 10+ hours of manual work * Success hinged on detailed task instructions and explicit state handling * Vague prompts, not flawed agents, were the biggest source of failure A great taster before the tour! Read it here [https://go.mlops.community/hxwn5z] [Read it here](https://go.mlops.community/hxwn5z) ## GenV: An Agentic Workflow for Actionable Insights from Google Meet Recordings Enough of boomers and millennials – meet GenV. GenV is a Python notebook that extracts structured insights from Google Meet recordings using Vertex AI’s Gemini model. It locates recent video files in Google Drive, uploads them to Google Cloud Storage, and runs them through a prompt-defined Pydantic schema. The output includes summaries, action items, and technical details. Insights are structured into fields like: * technical_insights – configs, tools, or code mentioned * projects_discussed – updates and progress notes * action_items – tasks with owners and deadlines Sadly, it still can’t explain what “skibidi” means. Read it here [https://go.mlops.community/n2gf02] [Read it here](https://go.mlops.community/n2gf02) ## Hidden Gems Working on something tricky or planning ahead? Here’s how we can help - just hit reply: * Custom workshops tailored to your company’s needs * Hiring? I know some quality folks looking for a new adventure * Want to connect with someone tackling similar problems? I can introduce you Thanks for reading, catch you next time! --- Source: https://aaif.live/newsletters/mlopscommunity/2025-05-15-agents