Writing
Notes
Architecture notes: patterns I design with, and lessons from building data and AI systems.
- Nov 2026
Agents and MCP: giving AI tools without giving away the keys
The Model Context Protocol standardises how agents reach tools and data. Governance is still your job.
- Nov 2026
BigQuery cost and performance: the levers that actually matter
Partitioning, clustering, the on-demand vs Editions decision, and the query habits that decide most of the bill.
- Nov 2026
Change data capture into BigQuery, without the surprises
Getting operational changes into the warehouse in minutes: capture, ordering, merging and the cost traps.
- Nov 2026
Delta Lake, Apache Iceberg and the open lakehouse
Open table formats let several engines share one copy of data. What they give you, and the decisions that matter more than the format.
- Nov 2026
RAG on enterprise data: an architecture checklist
Retrieval-augmented generation works when retrieval is treated as a data problem: permissions, freshness and evaluation first.
- Nov 2026
Governing a Databricks lakehouse with Unity Catalog
One catalog for tables, files, models and permissions, and the decisions to make before teams start creating things.
- Oct 2026
Domain Data Mesh Migration
Move from a central data platform to domain-owned data products without losing governance: tag, migrate in waves, keep policies shared.
- Oct 2026
Blind LLM Evaluation
Choose between models, prompts or documents with pre-registered criteria, anonymised inputs, repeated judging and order-swapped head-to-heads.
- Oct 2026
Governed Agent Operations
Let LLM agents act, but only through contracts, a write-ahead journal, risk tiers and a human gate, watched by a supervisor that can only read.
- Oct 2026
Cross-Cloud Lakehouse Disaster Recovery
Keep a governed, warm copy of a Delta lakehouse in a second cloud, with recovery objectives you can rehearse and measure.
- Oct 2026
Point-in-Time Correct Data
Make every past answer reproducible: record when data became known, query as of that moment, and stop the pipeline when data can't be trusted.