Job Description
Hi , I’m building a **sales AI copilot** and I need help building a **simple React web app**. The scope is very focused. There are only **two things** to build. ### 1) Real-time transcription (frontend) - Build a React web app - Capture microphone input from the browser - Connect to a real-time speech-to-text provider (Speechmatics or similar) - Use their real-time API (WebSocket-based) - Display the live transcript on the screen as it updates You don’t need to build speech models or backend logic — just consume the API and show the transcript. Speechmatics realtime docs (example): ### 2) Transcript chunking logic On top of the live transcript, implement **simple and reliable chunking logic**. What this means in practice: - The transcript text keeps growing continuously as speech is transcribed - The app maintains an internal “anchor” that represents **how much of the transcript has already been sent** - When I press a button or keyboard shortcut: - The app sends **only the portion of the transcript that appears after the anchor** - After sending: - The anchor is updated to the current end of the transcript - The next time I press send: - Only newly spoken text (since the last send) is included Important constraints: - Do not resend previously sent text - Avoid complex text diffing or heuristics - Prefer a simple, deterministic approach (e.g., tracking position/length) - The logic must be stable even during live, fast-changing transcripts The goal of chunking is to reduce latency and manual effort during live calls. ### Notes - No fancy UI needed - No database required - Everything can live in frontend state - Focus is correctness, speed, and reliability Let me know if this is something you can build and what your approach would be. Apply tot his job