Blog / Field notes
Practical notes on data systems, local AI, analytics, infrastructure, and the lessons that only show up after something runs.
01
Local AI / Agents / Ollama / Docker
Practical tools, patterns, and project ideas for running local models without turning the whole exercise into a GPU shopping contest.
All notes
Lessons: Jan-Jul 2026
5 min
The medallion diagram is simple. Overlapping files, unstable keys, snapshots, replay safety, and downstream expectations are where the architecture becomes real.
Lessons: Apr 2024-Aug 2026
5 min
End-to-end ownership changes the tradeoffs. You stop optimizing individual layers and start building for clarity, recoverability, and the next real operating decision.
From elsewhere
Articles, documentation, and references I found useful enough to keep. These links also appear alongside the posts they inform.
The useful starting point once local inference needs to become an application service instead of a terminal demo.
A practical guide to administering private compute without opening another public ingress path.
Boring, readable orchestration remains a strong default for small labs and application stacks.
The canonical Fabric framing for bronze, silver, and gold layers; useful context before adapting the pattern to real source behavior.
Reference material for MERGE mechanics, matched clauses, and the implementation details behind Silver-layer upserts.
A useful description of the space between raw data engineering and decision-ready analysis.