Mentat AI
Enterprise knowledge capture platform that uses AI-driven expert interviews and a knowledge graph engine to preserve institutional knowledge.
Problem
Companies lose critical institutional knowledge when experienced employees leave. Traditional documentation fails because tacit expertise — the kind that lives in people's heads — resists being written down.
Approach
Mentat uses AI-driven expert interviews to extract tacit knowledge through natural conversation. A knowledge graph engine connects insights across interviews, surfacing gaps and building a searchable institutional memory.
Started as a consumer conversational AI app for travel creators (iOS/Android), then pivoted to B2B enterprise knowledge capture based on customer discovery.
Key Metrics
- 90% pilot retention on the enterprise side
- 89% reduction in inference costs through custom model benchmarking across 6 vision models
- 19 backend services in production
- 1,733 tests in the test suite
Tech Stack Detail
Full-stack TypeScript/Python: Next.js 14 frontend with Bauhaus-inspired design system, FastAPI backend with PostgreSQL, Claude API for interview intelligence, Slack and GitHub integrations for knowledge delivery. Deployed on AWS with Docker.