Show HN: Find undervalued bikes near you with ML

Category: ai-ml

Tags: machine-learning, price-prediction, used-bikes, multimodal, bert

Score: 5.3/10 (Innovation: 6, Technical: 6, Documentation: 2, Utility: 4)

bike.broker is a web service that uses a multimodal deep learning model (BERT + tabular features) to predict fair prices for used bikes, incorporating custom parts and upgrades. It scrapes historical sold ads from Pinkbike and offers a chat interface for price queries. The project is interesting for its domain-specific application of ML to a real-world valuation problem.

Target audience: cyclists and used bike buyers/sellers

Repository: https://bike.broker

View on Hacker News