AI AgentsMay 18, 20263 min read
Building a Multi-Agent ESG Copilot
How I turned a brittle single-prompt ESG analyzer into a LangGraph multi-agent system with tools, validation loops, and human review.
Writing
Notes from shipping production AI
Essays on LLM systems, retrieval, agents, evaluation, and why production AI is mostly disciplined software engineering.
EngineeringApr 2, 20262 min read
Lessons from Shipping LLM Systems
Hard-won lessons from running LLM systems under real production load — routing, retries, timeouts, and the failure modes nobody warns you about.
LLM EvaluationMar 11, 20262 min read
Evaluation of AI Agents
Single-turn accuracy tells you almost nothing about an agent. How to evaluate trajectories, tool use, and failure recovery.
RAGFeb 20, 20262 min read
RAG Beyond Vector Databases
Vector search is table stakes. Real retrieval quality comes from hybrid search, graph traversal, reranking, and knowing when embeddings alone fail.
EngineeringJan 15, 20262 min read
Designing Reliable AI Systems
Reliability is a design property, not a patch. Fallbacks, degradation, guardrails, and observability for AI systems that stay up.
EngineeringDec 5, 20252 min read
Why Production AI is Mostly Software Engineering
The model is a small part of a production AI system. The rest is the software engineering that makes it trustworthy, observable, and maintainable.
EngineeringNov 8, 20252 min read
Building AI Products That Scale
Scaling AI products is about cost curves, latency budgets, caching, and evaluation velocity — not just throwing a bigger model at the problem.