Allsermon
Speech in. Structure out.
- Sole engineer
- Django · React Native · Expo · Next.js · PostgreSQL
- pipeline
- Real-timepipeline
- clients
- Mobile + webclients
The problem
People take notes during teachings and lose the thread. The useful artefact — a summary, the action points, a timeline — only exists afterwards, if at all.
The approach
Record, and get structure back in real time: full transcript, running summary, action points, timeline. Real-time rather than batch is the entire engineering problem — the pipeline has to stay ahead of a speaker who never pauses.
What I owned
Audio ingestion and the processing pipeline, the API, and both clients. Django and PostgreSQL on the server, React Native and Expo on mobile, Next.js on web.