Blog
Notes from the engineering bench.
Practical writing on shipping AI that holds up in production.
Grounding RAG: why retrieval isn't enough
Retrieval gets the right passages in front of the model. Validation is what keeps the answer honest. Here's how we use canonical segmentation, reranking and LLM-as-judge in production.
Read article →Agents that act — without losing the human
Autonomous agents are most useful when they do real work across real tools. They're only safe when people stay in control of the decisions that carry weight.
Read article →Useful computer vision starts with a webcam
You don't need specialist hardware to extract meaningful signal from movement. Standard video, transformer-based pose models and longitudinal tracking go a long way.
Read article →