DriveWorth
Used-car pricing for Pakistan, from a 1989 Mehran to this year's Fronx. DriveWorth prices a car from its condition, history and papers, points out where the story does not add up, and tells you what to say when you sit down to agree a price.
- Price model chosen from Ridge, Random Forest and Gradient Boosting by cross-validation: 96.7% accurate on 22,500 held-out listings, 98.3% within ±10%
- A four-step valuation with twenty plain questions, the eight CarSure section scores, or the full 135-item inspection checklist; an inspection report PDF fills itself in
- Every result has a price range, a factor-by-factor breakdown, red flags and a bargaining brief
- Deal Check gives a verdict on a seller's asking price with an opening offer and a walk-away price
- Five-year price forecast with sell or hold advice, side-by-side compare and a market overview
- A car-savvy chat assistant on the site and on WhatsApp that understands Roman Urdu questions and nicknames
- Cars and bikes, with a live price board on the home page
Built with
- Machine learning
- scikit-learn, pandas, Gradient Boosting
- Language AI
- Self-hosted LLM, WhatsApp Cloud API
- Backend
- Python, Flask, SQLite, PyMuPDF
- Frontend
- React, Vite