Skip to content
← Writing

IEEE DTU Vihaan 9.0 — Building an AI Healthcare Claims System Under Pressure

Our team built an end-to-end AI-powered healthcare claims processing system at Vihaan 9.0. We made it to the finals. Here's what we built, what broke, and what I learned.

·2 min read
HackathonAIHealthcareMulti-AgentRAGMachine Learning

Posting this a little late — almost 1.5 months after the fact — but this experience was too good not to share.

What We Built

At IEEE DTU Vihaan 9.0, our team built an AI-powered healthcare claims processing system designed to solve a real-world problem that impacts millions.

We engineered a complete end-to-end pipeline:

  • Vision AI for medical document extraction
  • RAG-based policy adjudication to match claims against insurance policy rules
  • Hybrid ML + LLM fraud detection to flag anomalous claims
  • A Mediator Agent that explains claim decisions in simple, human-readable language for patients

What made it special wasn't the tech stack — it was that this wasn't another hackathon PPT idea. We built an actual working system under intense time pressure.

Making the Finals

We made it to the finals. Didn't take home the trophy — but honestly, the learnings, feedback, and experience were far more valuable than any prize.

A big highlight was getting evaluated by industry leaders:

  • Gurpreet Singh — Senior Director, Capgemini
  • Sangh Gautam — CTO, UniVon
  • Tanya Raghuvanshi — AI at Airtel Africa | Co-founder, Peer Robotics
  • Pradeep Kumar Singh — CPTO, Aakash Educational Services Limited | ex-Amazon

Getting direct feedback from people operating at that level, on a system you built in under 24 hours, is rare.

The Real Challenge

One thing this hackathon made very clear: the real challenge isn't writing code quickly. It's making the right decisions when systems fail, time is running out, and the team still has to deliver.

There were moments where we had to completely rethink parts of the fraud detection pipeline — multiple times — before it finally started making practical sense. Every iteration improved the product. The pressure didn't slow us down; it sharpened our thinking.

Why It Felt Meaningful

Healthcare fraud is still a massive unsolved challenge. Building something that can autonomously detect anomalies using AI, ML, and clinical reasoning — even in a hackathon context — felt like it pointed at a real problem worth solving.

Thanks

Huge thanks to our mentors Aakash Gupta (SDE-2, Expedia Group) and Yogita Khurana (SDE-2, Microsoft) for their guidance and feedback throughout the journey.

And couldn't have done this without the team:

  • Sachin Manral
  • Simran Kukreja
  • Vaibhav Yadav

Really grateful for the experience — the pressure, the sleepless hours, and every lesson that came with it.

Comments