Bidlyzer.ai

• 

  Making government contract frauds very hard to commit

Inspiration

I’m a huge fan of the YC CEO’s call for bold AI startups. Watching Watch Dogs and Ozark sparked my fascination with fraud, and now I’ve turned that obsession into a research paper and a working prototype.

What it does

Bidlyzer AI scans government contract PDFs, flags compliance risks, cross‑checks entities against SAM.gov, USASpending, and the ICIJ leaks, and spits out a ready‑to‑use investigation report.

How we built it

We stitched together a FastAPI backend, a Vite + React frontend, and a LangGraph pipeline of four lightweight agents. All the heavy‑lifting LLM work is mocked for the demo, keeping the repo lightweight and hackathon‑ready.

Challenges we ran into

Reading 93 research papers and curating our own fraud‑detection dataset was a marathon. Translating academic insights into runnable code took a lot of trial‑and‑error.

Accomplishments that we're proud of

We turned those research hurdles into a functional demo, built a sleek UI, and proved the pipeline can flag every known fraud case in our test set.

What we learned

Going from zero to one is exhilarating just like Peter Thiel says, “It’s the most fun you can have building something that didn’t exist before.”

What's next for Bidlyzer.ai

We’re gearing up to turn this prototype into the next unicorn in the startup leagues. Stay tuned!

Built With

• a-large?scale-llm-backend
• built-with?python?+?fastapi
• langgraph
• sqlite
• typescript?+?react?+?vite