SuperCall

Inspiration

SuperCall was inspired by real pain points in venture capital workflows—especially the difficulty of tracking scattered deal information across emails, chats, and meetings. The idea came from firsthand experience with teams struggling to maintain context and prepare effectively for high-stakes conversations.

What it does

SuperCall is an AI-powered real-time meeting intelligence tool that ingests historical data (emails, notes, chats) and provides live insights during conversations. It connects ongoing discussions with past context, highlights key signals (commitments, risks, follow-ups), and generates structured summaries after meetings.

How we built it

We built SuperCall using a multi-stage AI pipeline:

Data ingestion from multiple sources (email, documents, chat)
Chunking and embedding into a vector database for semantic retrieval
Real-time speech-to-text transcription with speaker diarization
Semantic matching between live conversation and historical data
LLM-powered reasoning layer to generate insights and summaries

Challenges we ran into

Ensuring accurate semantic matching between live speech and historical context
Designing a system that generalizes beyond VC to other domains

Accomplishments that we're proud of

Built a working real-time AI meeting assistant
Delivered actionable insights (not just transcription)
Designed a flexible system applicable to multiple industries
Created a live demo and deployable prototype

What we learned

Conversational AI is extremely valuable in high-stakes meetings where context and memory are critical. Real-time augmentation of human conversations can significantly improve decision-making, preparation, and follow-ups.