AI / ML2024-05
AI Meeting Assistant & Voice Automation
Manual note-taking in meetings was eating up hours every week, and important action items kept falling through the cracks. I built a meeting intelligence platform that captures live audio via WebRTC, transcribes in real-time, summarizes key points, and automatically assigns action items with calendar integration.
Key Highlights
- For sensitive environments, there's a fully offline mode using Whisper.cpp and local LLMs — no data leaves the room
- Even in noisy conditions it hits 85% accuracy
- Teams using it save an average of 5 hours per week on meeting admin
Technology Stack
AI
WhisperWhisper.cppGPT-4Local LLM
Infrastructure
WebRTC
Core
ReactPythonC++