SalesIQ

SalesIQ

Inspiration

One of our team members had a lot of experience with digital marketing and the challenges associated with it. Giving a product demonstration can be a very stressful experience. We wanted to build an AI coach that could analyze their performance in real-time and give them actionable feedback without breaking their flow.

What it does

SalesIQ captures live sales calls and uses computer vision plus Claude AI to analyze body language, tone, and conversation flow. It gives real-time suggestions during the call and generates a detailed post-call breakdown showing what worked and what didn't.

How we built it

We used LiveKit for real-time video streaming and WebRTC capture, then built a pipeline that feeds frames to computer vision models for emotion detection. The transcript and visual data get sent to Claude API which generates contextual advice on the fly. Everything gets stored in Firebase with Firestore for metadata and Storage for video files.

Challenges we ran into

Using livekit to record real time audio and relay insights back to the frontend in real time was definitely a learning curve but once we figured it out it started making sense.

Accomplishments we're proud of

We got the entire real-time pipeline working end to end. Video capture, emotion analysis, Claude integration, and Firebase storage all communicate seamlessly. The dashboard actually pulls real call data and displays it in a way that makes sense for sales coaching.

What we learned

Real-time AI is way harder than batch processing. Coordinating multiple APIs while keeping latency low taught us a lot about asynchronous programming and optimizing data flow.

What's next

We want to improve conversation pattern recognition so it can identify successful sales techniques across multiple calls. Integration with actual video conferencing platforms like Zoom would make this immediately usable.