The Story
This dashboard started as a PhD-budget side hustle: buying damaged cars at insurance auctions, fixing them, and selling them on. The full write-up is a three-part series on prasanthsasikumar.com.
Scraping three years of NZ auction listings
How a daily crawler and GitHub Actions quietly built a 660,000-record public dataset.
What 1,100 days of salvage auctions reveal
Price trends, damage patterns, and how this site serves an API with zero servers.
Predicting auction prices with XGBoost
Turning the logged history into a practical "is this a good buy?" estimator.
Try the price predictor
A Streamlit app wrapping the XGBoost model: paste a listing's details, get a sensible price estimate.





