BonVivant Blog
Behind the Build
How we built an AI-powered discovery platform — enrichment pipelines, RAG patterns, and the engineering decisions behind BonVivant.
The LLM Judge That Says "Check This," Not "This Is Wrong"
How we use an LLM-as-Judge as a publish gate for hand-written pages — and why its factual-accuracy score is a verification queue, not a verdict.
LLM-as-Judge with Feedback Loops: How We Got an 85% Rejection Rate Down to 38%
Building a self-correcting editorial pipeline for venue content at BonVivant.
Building BonVivant: An AI Discovery Platform That Actually Has Taste
Why we built an opinionated alternative to Google Maps reviews — and what we learned trying to teach a machine to care about lamb shoulder.