emmer-stui — AI-Powered Conference Call Assistant

Inspiration

Traditional conference calls are often inefficient, hard to follow, and time-consuming to summarize. We wanted to create an AI assistant that can actively join phone calls, listen, understand, summarize, and take action—just like a real teammate.
Our goal: make every meeting more productive, more focused, and less painful.

What It Does

emmer-stui is an AI-powered conference call system that:

• Automatically joins phone calls and listens in real time
  
• Provides live speech-to-text transcription with multilingual support
  
• Generates meeting summaries, key points, and action items instantly
  
• Speaks in real time to answer questions or provide information
  
• Syncs action items to tools like Notion, Trello, Slack, or email
  
• Analyzes conversation flow, speaking time, and discussion topics
  
• Works with phone dial-in, online meetings, and mobile devices
  

In short: it transforms meetings into organized, actionable insights—automatically.

How We Built It

We built emmer-stui with a blend of AI, telephony, and real-time processing technologies:

• AI Models: Large Language Models for real-time understanding, summarization, and Q&A
  
• Speech Recognition & TTS: WebRTC + speech APIs for low-latency interaction
  
• Backend: Node.js + Express for call control, AI orchestration, and APIs
  
• Frontend: React + Tailwind for the meeting dashboard and control panel
  
• Telephony Integration: Twilio / Vonage for call routing and phone participation
  
• Database: PostgreSQL + Redis for caching and high-speed data handling
  
• Deployment: Docker + Vercel / AWS for scalable, containerized deployment
  

The result is an AI assistant that can actively participate in meetings.

Challenges We Ran Into

• Maintaining low latency between speech recognition and AI responses
  
• Handling noisy audio and overlapping voices from real phone calls
  
• Ensuring the AI remains context-aware across long, dynamic discussions
  
• Managing telephony API callback events and exceptions
  
• Achieving stable, structured action-item extraction in multiple languages
  

Accomplishments That We’re Proud Of

• Successfully implemented an AI assistant that can join real phone calls
  
• Achieved live listening, speaking, summarizing, and question answering
  
• Built a clean, intuitive dashboard for monitoring and exporting insights
  
• Completed a full audio → understanding → action pipeline within hackathon time
  
• Received strong initial feedback from test users praising meeting efficiency
  

What We Learned

• Real-time voice AI systems require micro-optimization at every step
  
• LLMs perform best with structured prompts and context-aware chaining
  
• Telephony systems involve many edge cases (busy lines, disconnects, delays)
  
• Effective teamwork and rapid prototyping are crucial in hackathons
  
• The potential for AI-assisted communication is far larger than expected
  

What's Next for emmer-stui

We plan to expand emmer-stui with:

• AI that can proactively ask clarifying questions and prevent misunderstandings
  
• Cross-meeting memory and organizational knowledge base
  
• Live subtitles and instant translation for multi-language meetings
  
• Deeper integration with corporate ecosystems (Google Workspace, Microsoft Teams, Feishu)
  
• Automated meeting analytics dashboards and visual reports
  
• A mobile app and Chrome extension for instant access anywhere
  

Our vision: make emmer-stui the smartest teammate in every meeting.

Built With

• claude
• mysql
• python