Insights

Notes from the build.

How we engineer production AI for regulated industries — evals, retrieval, agents, and sovereignty. Written for the people who'll run it.

• Vendor-neutral platform · 14 June 2026 · 2 min read

  Self-hosted vs Bedrock vs Azure OpenAI for sovereign workloads in Australia

  A practical decision framework for where your model actually runs when data residency and sovereignty are hard constraints — and how to keep the choice reversible.

• RAG & retrieval · 12 June 2026 · 3 min read

  Shipping a RAG system that survives an APRA audit

  Grounding, inline citations, retrieval evals, and a refusal path — the engineering that makes a retrieval system answer to an auditor, not just a demo.

• LLMOps & reliability · 9 June 2026 · 3 min read

  Evals before vibes: the gates we run on every LLM release

  How we stop 'looked good in the demo' from reaching production — the offline eval suite, the regression gate, and the tracing we wire in from day one.