KC Fireflies Secure Meeting Bot
Containerized Multi-Platform Meeting Recording & Knowledge Management
Architected and developed a containerized microservice solution for automated meeting recording, transcription, and knowledge artifact capture across Microsoft Teams, Google Meet, and Zoom, with seamless integration into enterprise knowledge management systems.
Client
Luxottica
Completion
4 months
Category
Automation & AI
Situation
The organization faced challenges with disjointed platforms and high administrative overhead when capturing and publishing critical knowledge artifacts from meetings. Traditional methods required manual intervention, lacked integration with knowledge management systems, and created compliance risks for sensitive meetings. The need for a cost-effective, secure, and automated solution was critical.
Task
Design and implement a containerized, automated meeting bot solution that can join and record meetings across multiple platforms, capture transcripts and summaries, integrate with enterprise knowledge systems, and provide a secure alternative for sensitive meetings that cannot use external AI processing.
Action
→Architected a containerized microservice using TypeScript, Node.js, and Playwright for browser automation to join and record meetings across Microsoft Teams, Google Meet, and Zoom
→Deployed the solution on Azure Container Instances (ACI) using a custom Ubuntu container with Chromium and Playwright, enabling cloud-native, scalable meeting recording
→Developed six RESTful API endpoints to queue meeting recordings with parameters like user ID and meeting URL, implementing a single job execution model per container for reliability
→Integrated with Azure Blob Storage for secure recording storage in MP4 format with custom codec for stable playback in SharePoint
→Implemented webhook triggers upon recording completion to initiate event-driven workflows in Power Automate and N8N for post-processing and transcription
→Designed integration with Knowledge Central and CIAO Planner, automatically routing AI-detected action items into Planner tasks with unique Meeting IDs (KCID) for traceability
→Created an Omnibar search workflow enabling users to quickly retrieve meeting recordings and transcripts using KCIDs or SRT numbers
→Developed a self-contained solution using Kasm Workplaces with Ubuntu Docker container for secure meeting recording without consuming local user resources
→Integrated Firefly AI for transcription and summary extraction, optimizing license usage to a single seat for cost efficiency
→Planned integration with Open Web UI for automated notes and AI-driven insights, with local LLM proxy architecture for sensitive queries
Results
✓Achieved operational cost of approximately $0.30 per 90-minute meeting using Azure Container Instances, significantly lower than traditional VM-based deployments
✓Realized substantial cost savings by utilizing Firefly AI with only one seat license instead of multiple individual licenses (estimated $228/year per seat savings)
✓Eliminated the need for expensive Microsoft Copilot licenses by providing auto-transcription capabilities through the custom containerized solution
✓Streamlined knowledge management workflow ensuring action items are context-rich and traceable, reducing back-and-forth communication and accelerating task progress
✓Enabled rapid knowledge retrieval through Omnibar search workflow, dramatically speeding up access to documentation and meeting context
✓Enhanced compliance and security by utilizing internally managed infrastructure, reducing reliance on external vendors for core processing
✓Provided a secure meeting bot solution for sensitive information that cannot be processed by external AI services
✓Established a scalable, multi-platform solution supporting Teams, Google Meet, and Zoom with consistent recording quality and format