Writing
Technical notes on production AI systems, LLM infrastructure, and engineering reliability.
Evaluation-driven LLM systems in production2026-05-17
How I used DSPy signatures, self-consistency, retrieval feedback loops, and MLflow experiment tracking to build a repeatable optimization system for production LLM workflows.
LLMsEvalsDSPyMLflowProduction AI
Why we avoided ML for behavioral anomaly detection2025-06-01
At MVP scale, per-household ML models would be underfit and hard to interpret. Why statistical baselines were the right call for a safety-critical eldercare system.
Anomaly DetectionProduction AISystems Design