quackDeal

Inspiration Every sales rep has felt it — you walked out of a meeting thinking it went great, then the deal goes cold and you have no idea why. We wanted to fix that feeling with real data.

What it does quackDeal is an AI-powered sales intelligence platform. It analyzes sales meetings — transcripts, screenshots, or live recordings — and instantly scores deal health based on objections, sentiment, and commitment signals. Sales reps get real-time alerts when deals go cold, AI-generated follow-up emails, and a Pipeline AI assistant they can ask in plain English: "which deals are at risk?" or "what did Acme say about budget?"

How we built it Claude AI for meeting analysis and insights. Snowflake as the data warehouse. Databricks for the pipeline. Firebase for auth. React and Node.js/Express for the frontend and backend. Four engineers, 24 hours, one shared screen.

Challenges we ran into Claude's API doesn't support audio files — we pivoted to Web Speech API for live recording. Snowflake VARIANT columns required careful JSON handling. Git history had to be scrubbed after secrets were accidentally committed. We lost sleep so you don't lose deals.

Accomplishments that we're proud of A fully working product in 24 hours. 100-record seeded dataset across 8 industries. Real-time deal health scoring, AI chat, insights generation, smart reminders, and follow-up emails — all live and connected end to end.

What we learned Ship fast, fix faster. Read the API docs before assuming audio works. And always add .env to .gitignore before your first commit.

What's next for quackDeal CRM integrations (Salesforce, HubSpot). Real-time meeting transcription via Whisper. Team dashboards for sales managers. Predictive win/loss modeling trained on historical deal data. And maybe some sleep.

Built With

• claude
• databricks
• express.js
• firebase
• javascript
• react
• snowflake