AI products · Local prototype
ContextCoach
Practice a conversation, then inspect the feedback.
A local English conversation-practice prototype with streamed responses, browser speech, contextual feedback and a review of the session.
The question
Practice needs clear turn-taking and useful feedback, including predictable behavior when a response is interrupted or a provider fails.
How it works
- Use bidirectional WebSockets for response deltas, feedback, cancellation and reset.
- Keep provider credentials on the server and provide a deterministic mock fallback.
- Offer evidence-linked session review and opt-in browser-local progress summaries without persisting transcript text in those summaries.
Evidence & results
- Local mock-mode tests cover streaming state, cancellation, review and telemetry paths.
- Retained desktop/mobile captures show the working practice and session-review interface.
- A local timing study measures mock pacing, with its environment and limits recorded.
Scope & next steps
- No demonstrated learning improvement, validated proficiency score or real participant study.
- In-memory sessions; no verified production authentication, persistence or AWS deployment.
- Mock pacing improvements are not faster live-model inference.
Go one level deeper.
Ask about the implementation, tradeoffs, or my contribution.
Explore with the portfolio guide