Writing
Field notes on AI engineering, agentic systems, data architecture, and the craft of building at scale.
MCP is the missing standard that lets AI agents actually do work in the enterprise — safely, auditably, and without a bespoke integration for every tool.
Snowflake cost optimization isn't about turning things off — it's about warehouse discipline, query hygiene, and modelling. Here's the playbook that reliably works.
Why the next generation of data platforms won't just move data — they'll reason about it, fix themselves, and do the work autonomously.
The lakehouse-versus-warehouse debate generates more heat than light. Here's the practical framework I use to make the call for real enterprise platforms.
AI is changing what data engineers do, not whether they're needed. Here's how I lead a team through that shift without losing morale, quality, or the plot.
Most RAG demos are magic. Most RAG products are disappointing. The gap is almost never the model — it's the retrieval.
The move from senior data engineer to engineering lead isn't more of the same skills — it's a different job. Here's what actually mattered, fifteen years in.
How replacing hand-written pipelines with declarative configuration turned days of work into hours — and made the whole team faster.