Inspiration
Modern discussions and meetings move quickly. Whether in sales calls, product demos, or strategic planning, participants often struggle to keep up with information that is fast, high pressure, and information heavy. Many key points are lost during real-time discussions, leading to inefficient communication and missed opportunities.
This inspired us to build BargAI, a real-time AI meeting assistant that helps professionals process live conversations and turn them into actionable insights.
What it does
BargAI transforms web-based meetings into structured, intelligent discussions by: Listening to live conversations through Google Meet or other conference platforms Using speech-to-text processing to transcribe in real time Sending conversation data to an AI model (Gemini 2.5 Flash) for instant analysis Displaying contextual insights, summaries, and key takeaways through a clean interface BargAI helps users stay focused during meetings, track important points, and make better decisions based on what is being discussed.
How we built it
The system connects web conferencing, data processing, and AI reasoning into one flow: Web-Based Platform Integration: Selenium captures and parses live captions or transcripts from Google Meet, Zoom, or Teams. Data Processing Layer: Python compiles the live transcript and sends it to Gemini for reasoning. Language Model: Gemini 2.5 Flash interprets context, retrieves relevant information, and formulates insights. User Interface: Built using HTML and JavaScript to visualize insights and summaries. Database Connection: A read-only database (such as company data or meeting archives) is optionally referenced through App Script or Webhooks to provide context without modifying any stored data.
Teck Stack
Frontend: JavaScript, HTML Audio Capture: Web-based conference platforms (Google Meet, Zoom, Teams) Speech-to-Text: Platform caption systems (Google Meet captions) Backend: Python, Selenium Language Model: Gemini 2.5 Flash Web Servers: App Script and Webhook
Challenges We Faced
Handling simultaneous microphone and system audio to capture both speakers Ensuring transcription and AI response latency stayed under three seconds Designing a live overlay that was simple, readable, and not intrusive Managing AI reliability when working with dynamic, real-time data
Accomplishments that we're proud of
Created a functional prototype that transcribes, analyzes, and responds in real time Built an integrated Electron interface with Gemini reasoning Designed a working architecture capable of fetching contextual data from local files Demonstrated live insight generation with a clear, intuitive interface
What we learned
How to synchronize audio, transcription, and AI reasoning into a single real-time workflow How to design readable, low-latency UI systems for live contexts How to control AI output for context relevance and accuracy The importance of user experience in AI tools designed for communication settings
What's next for BargAI
Automatic generation of meeting summaries and action items Multi-language support for global collaboration Emotion and tone detection for improved context understanding Integration with CRM systems and enterprise collaboration tools
Log in or sign up for Devpost to join the conversation.