Writing about AI engineering, quality systems, and the practical craft of building reliable software in the age of large language models.
Six months into our team's GenAI journey, here are the hard-won lessons from 60 days of building, evaluating, and shipping AI-powered features in production.
Building a fully local PDF question-answering system using DeepSeek-R1, Ollama, and a simple vector store — no API keys, no data leaving your machine.
Impressions, announcements, and the conversations that actually mattered at OpenAI's first London DevDay.
A practical guide to running open-weight language models locally using Ollama, Docker Compose, and a simple web UI — from first install to production-like setup.