MeetSync AI

Inspiration

Organizations lose critical information from meetings due to fragmented platforms and lack of searchable memory. MeetSync AI is built to create a unified conversational intelligence layer that makes meeting data discoverable and actionable across all platforms.

What it does

Multi-Platform Integration: Connect Google Meet, Zoom, Microsoft Teams, Slack, and more AI-Powered Search: Natural language queries with hybrid search (keyword + semantic) Intelligent Summarization: Automatic meeting summaries, action items, and decision tracking Calendar Integration: Sync with Google Calendar, schedule meetings, set reminders Multi-Channel Notifications: Email, SMS, WhatsApp, and push notifications Collaboration Network: Visualize and analyze team collaboration patterns Advanced Analytics: Meeting trends, sentiment analysis, and participation metrics Timeline Visualization: Track feature decisions and meeting themes over time

How we built it

The backend uses Flask with Vertex AI for natural language processing and RAG. Fivetran connectors ingest data from meeting platforms into BigQuery for structured storage and Cloud Storage for transcripts. Elasticsearch provides hybrid search combining keyword matching with semantic embeddings. Redis caches frequent queries, and the frontend uses Plotly for interactive visualizations.

Challenges we ran into

Implementing hybrid search required balancing keyword precision with semantic understanding. Handling huge data sync across multiple platforms with different API structures needed careful schema normalization. Managing state between demo and production modes while maintaining feature parity required conditional logic throughout the codebase.

Accomplishments that we're proud of

Created a working system that processes meeting transcripts, generates actionable summaries and enables natural language queries. Built interactive network visualizations showing collaboration patterns and implemented timeline views tracking decision evolution across quarters.

What we learned

Hybrid search architectures require tuning weights between semantic and keyword components. Data ingestion from heterogeneous sources demands flexible parsing strategies. Caching strategies significantly impact query performance in conversational AI systems.

What's next for MeetSync AI

Implement meeting transcription with speaker diarization. Add automated action item tracking with assignment notifications. Expand platform support to include Notion, Asana and Trello for complete project context. Deploy predictive analytics for meeting effectiveness and scheduling optimization.

Built With

• apscheduler
• axios
• binary-parsers
• cosinesimilarity
• docker
• elasticsearch
• fivetransdk
• flask
• github
• google-bigquery
• google-calendar
• google-cloud
• google-cloud-run
• json
• jwt
• plotly
• python
• redis
• sendgridapi
• snowballstemming
• tailwind
• twiliorestapi
• vanilla
• vertexai
• xml