ChurnGuard

Inspiration

What it does

ChurnGuard is an AI agent that predicts customer churn risk from uploaded data (e.g., tenure, monthly charges, contract type). User inputs values gets churn probability (e.g., 80.57%) + risk level (Low/Medium/High) via live API. No manual modeling fully automated analysis and deployment.

How we built it

AI: Integrated Anthropic Claude (or Gemini) to analyze CSV, build model, generate predictions. Deployment: Zerve AI notebook for data processing + FastAPI endpoint for live prediction (/predict). Data flow: Upload CSV AI processes model saved FastAPI serves predictions.

Challenges we ran into

CORS errors: Browser blocked cross-origin requests added app.use(cors()) in Express. Model file not found during deploy ensured pickle dump/load paths matched exactly. API endpoint discovery Zerve docs were unclear; trial-and-error to find /predict route. Limited time: Balancing MVP features with polish under 48-hour crunch.

Accomplishments that we're proud of

Built end-to-end AI agent from idea to live API in under 2 days. Achieved real prediction accuracy (80%+ churn detection on sample data).

What we learned

FastAPI + Zerve = rapid deployment cut weeks of work to hours. Hackathons teach speed + iteration—better than months of solo planning.

What's next for ChurnGuard

Submit to more hackathons & apply learnings to real fintech products. Add real-time monitoring dashboard (trends, alerts).

Built With

• https://churn-predict-api.hub.zerve.cloud/predict