Deep dives into projects and engineering decisions.
How we built a VLM-based safety detection agent for manufacturing environments — covering Enriched Input, Two-Stage Reasoning, and PPE mode.
Read post ↗How adding graph database relationships to a document RAG pipeline improved internal benchmark performance by ~15%.
Read post ↗An end-to-end evaluation agent that scores NL2SQL outputs and feeds structured improvement signals back into the development cycle.
Read post ↗A hackathon-winning agent that analyzes UX guidelines and in-progress screens to detect guideline violations automatically.
Read post ↗An Electron-based local AI platform where users install task-specific plugins — meeting summarization, file search, and more — without sending data to the cloud.
Read post ↗