Selected work

AI / RAG · 2026

Meeting Intelligence System

A grounded AI workspace for turning meeting transcripts into useful, auditable knowledge. It combines retrieval, structured extraction, and a deliberate abstention path so the product does not present unsupported answers as fact.

What it does

01

Grounded answers

Questions are answered from retrieved transcript chunks, with citations that let a user jump straight to the supporting moment in the meeting.

02

Evidence before fluency

When retrieval confidence is below a threshold, the app answers “not discussed” without calling the LLM, reducing hallucination and unnecessary cost.

03

Useful by default

Ingestion extracts decisions, action items, and key topics so each meeting is valuable before anyone types a question.

Built with

PythonFastAPIClaudeVoyage embeddingsChromaDBReact

Want to talk through the thinking behind it?

I'm always interested in useful product and platform conversations.

Get in touch