Every year or so someone publishes a take claiming that feature engineering is dead — AutoML will do it, foundation models will make it irrelevant, end-to-end deep learning will learn the representations directly from raw data. And every year, the teams doing serious ML work on enterprise tabular data are still
Nine months of building a context system for AI-assisted development had produced one clear lesson: the problem is not storage, it's retrieval. Every approach I had tried — flat markdown, structured markdown, a PostgreSQL schema with a JSONB blob — was a better or worse answer to the question "
DAIS 2024 had multiple sessions on efficient model training, distillation, and domain-specific fine-tuning. I want to pull on the thread that connects all of them, because it points to a conclusion that's going to be uncomfortable for anyone who has been telling their organization they need the biggest,
There's a quiet exhaustion settling into data teams that have spent 2024 chasing GenAI use cases that haven't shipped, building RAG pipelines that don't quite work well enough, and defending to stakeholders why the AI chatbot isn't ready yet. Meanwhile, the gradient
Databricks highlighted cost governance at DAIS 2024 — GPU efficiency, cost monitoring, inference optimization. It's worth talking about why they felt they needed to say this out loud. The reason is that enterprise teams building GenAI systems in 2024 are consistently surprised by the cost. Not pleasantly. The Surprise
Every system has a wall — the point where its design assumptions stop holding and the cracks start showing. For markdown-as-memory, the wall appeared around month nine. It wasn't a dramatic failure. It was an accumulation of small frictions that added up to a clear signal: the format had
At DAIS 2024, Unity Catalog was extended to govern not just data tables but models, features, vector indexes, and lineage that spans the full machine learning lifecycle. I've spent the last few months integrating this more deeply into real projects, and the expanded UC story is genuinely changing
Three months after DAIS and I've now been in enough post-conference retrospectives to see the same pattern play out. The team that attended Summit comes back energized. They have three to five AI use cases they want to build. They get budget approval for a proof of concept.
The thing nobody says loudly at AI conferences: most GenAI application failures are retrieval failures. The model itself — the LLM doing the generation — is usually not the problem. The problem is that the model received the wrong context, incomplete context, or no relevant context at all, and generated confidently wrong
There's a point where a workaround becomes a problem in its own right. I crossed that line somewhere around month eight, when I noticed that my weekly calendar had a recurring block labeled "context maintenance" that was taking longer than the original task it was supposed
The natural question, given that I'm building LLM-powered data tooling at the moment: can you use a language model to generate a Great Expectations suite? Profile a dataset, describe the domain, and have the model produce a meaningful suite that doesn't require hours of manual expectation
Two months out from DAIS, I've had time to actually use Mosaic AI rather than just watch the demos. The unified branding that Databricks put on this product family at Summit — training, serving, vector search, agent framework, tools catalog — was clearly designed to compete with the fragmented GenAI