Blog
Engineering notes
Practitioner takes on AI agents, RAG, LLM orchestration, cloud and FinOps, platform architecture, and security, drawn from production work.
RAG vs fine-tuning: when to use which
When should you reach for retrieval-augmented generation, when for fine-tuning, and when for both? A practical decision framework from production work.
How to build a production RAG pipeline
A walkthrough of the stages that decide whether a RAG system works in production: ingestion, chunking, retrieval, reranking, grounding, and evaluation.
LLM cost optimisation: how to cut inference costs
The levers that actually reduce LLM inference spend: model routing, prompt and response caching, token budgets, and treating model cost as a FinOps discipline.
How to prevent prompt injection in LLM apps
Prompt injection cannot be solved with a single filter. A layered defence for LLM apps and agents, from input handling to tool design and least-privilege access.
What is LLM orchestration?
LLM orchestration coordinates models, prompts, tools, and state across a request. What it covers, when plain code is enough, and when an orchestration layer earns its keep.
Event-driven vs microservices: when to use which
Event-driven architecture and microservices are often framed as a choice. They are not. How the two relate, the coupling tradeoffs, and when each pays off.