Series

Building Herald — Private AI From Scratch

End-to-end build log of Herald — a fully local, private AI assistant running over 20 years of personal documents. No cloud APIs, no data leaving the machine. Covers RAG pipelines, vector databases, local LLMs, voice interfaces, agentic AI, and everything in between.

NVIDIA NIM vs Ollama — Which Should You Choose for Local LLM Deployment?
A practical comparison from someone who has run both in production
Mar 12, 20269 min read23
In, Store, Retrieve: The Three Places RAG Quietly Fails — And How We Fixed Each One
What five months of building a private AI document intelligence system taught me about retrieval quality, LLM reliability, and the gap between "it works" and "it works well"
Mar 17, 202618 min read15

Command Palette