Paawan Desai

Paawan Desai

he/him

I build AI agents, and the tools that test them and keep them safe. Gauntlet AI Cohort 7, UT Austin ECE '25.

← All projects

Prediction-Market Aggregator

What it does

Pulls markets and order books from Kalshi and Polymarket and normalizes them into one venue-independent model. It classifies which contracts are economically the same bet, with a written definition and a contract diff, and simulates routing a hypothetical order with a plain-English explanation. No real orders are ever placed.

From two venues to one routing decision
  1. Kalshi + Polymarketmarkets, order books
  2. Fetch every 2 minscheduler writes to Postgres
  3. Normalizeone venue-independent model
  4. Match contractssame bet, with a contract diff
  5. Simulate routingwhere an order would fill

Engineering notes

  • A scheduler collects a fetch round every 2 minutes. Pages read only from the database, so serving a page makes zero venue API calls, and a test enforces that.
  • Pairs come from a pinned round so the demo stays stable, and every price shows its age.

Tools

Go, Postgres, Docker, Google Cloud Run, Cloud SQL