Field notes
Write-ups from real projects: AWS architecture, infrastructure-as-code, platform engineering and applied generative AI.
Evaluating RAG with context precision, recall and an LLM judge
GenAI · · 1 min readIf you cannot measure retrieval, every prompt tweak is a guess. Here is the evaluation loop our team built, and what it taught us.
Read the post →- GenAI
Evaluating RAG with context precision, recall and an LLM judge
If you cannot measure retrieval, every prompt tweak is a guess. Here is the evaluation loop our team built, and what it taught us.
- AWS
Landing a multi-account AWS organisation, guardrails first
Control Tower, SCPs and a security baseline that teams can live with.
- Architecture
Running a DCIM platform on AWS
Lessons from moving data centre management software into the cloud across several sites.
- GenAI
Exposing a NestJS agent platform over MCP
One server, four agents, and the interface choices that kept it simple.
- GenAI
Config-driven agents: multi-agent simulation with YAML and Bedrock
Defining agents as data so scenarios change without code changes.
- Engineering
Why this site runs on Astro, S3 and CloudFront
Markdown content collections, static output and a hosting bill measured in cents.
Get new posts by RSS
No newsletter, no tracking — just a feed.