Why this exists

I'm a senior data engineer. I've spent years building production data pipelines, and the last few of those increasingly on AI/ML workloads — embedding pipelines, RAG infrastructure, vector store ops, eval data engineering, the unglamorous plumbing that makes “production AI” actually work.

There's a gap. Data engineering newsletters cover Spark, dbt, and Airflow. AI newsletters cover models, benchmarks, and product launches. Almost nothing covers the intersection — the actual engineering work of making AI run reliably in production.

That's what this newsletter is. One email every Tuesday. Curated links to what mattered that week, with commentary on why it matters to someone running AI workloads in production.

“Production AI is a data engineering problem.”

What you'll get

• →Tooling evaluations focused on the data/AI boundary
• →Production stories from teams running AI at scale
• →Cost and pricing shifts in embedding, retrieval, and storage
• →Postmortems worth reading — real failures, real lessons
• →Practical architecture patterns, not academic papers
• →Curated signal — every link is read before it's included

What you won't get

• ×Republished content — we link to sources and quote sparingly
• ×Daily emails — one per week, Tuesday morning
• ×Sponsorship-driven editorial — sponsors don't pick the stories
• ×Generic AI takes — every story has a data engineering angle

If something in an issue doesn't land for you, hit reply. A real human reads every reply. That's the deal.