Nudge is an intelligent meeting assistant that keeps discussions on track. Powered by AI and real-time speech analysis, Nudge helps teams maximize productivity and minimize wasted time.
nwHacks 2025 — LLM Track Finalist
Meetings often derail from their core objectives, wasting valuable company resources and employee time. Nudge solves this problem by:
- Real-time topic detection — Analyzes speech to determine if discussions stay on-topic
- Smart notifications — Issues warnings when the conversation drifts
- Productivity insights — Provides statistical analysis to improve future meetings
- Structured planning — Schedule meetings, define time blocks, and set specific topics
With Nudge, organizations save money while employees reclaim productivity.
Nudge combines cutting-edge technologies for efficient meeting analysis:
- Frontend: React + TypeScript with real-time WebRTC video streaming
- Backend: Flask with WebSocket support for audio processing
- AI Model: Locally embedded Llama 3 for sentence generation and topic analysis
- Vector Processing: Sentence Transformers for semantic representations
- Similarity Matching: Cosine similarity algorithms for on-topic detection
How it works:
- Audio batches are streamed via WebRTC to the Flask backend
- Llama 3 generates candidate sentences likely to appear in the meeting
- Sentence Transformer converts text to mathematical vectors
- Cosine similarity compares vectors to determine topic relevance
- Real-time feedback keeps discussions focused
Quick Demo — YouTube: On-Topic Check Demo
(Note: Demo focuses on core feature; schedule and extended features added after submission)
Project Page — Devpost Submission
This project is licensed under the MIT License