AI-driven used-car valuation prototype created for the first Greek AI Hackathon, powered by ACEin at the Athens University of Economics and Business.
- Project period: November 2023 – May 2024
- Status: Completed hackathon prototype; the original service is no longer live
Watch the Worthify platform showcase
Worthify explored how machine learning could make used-car pricing more transparent for individuals, dealers and insurers. A user supplied a vehicle's characteristics and the prototype returned an estimated market value through a guided web experience.
The longer-term concept also included forecasting: combining historical prices with market signals to explore how the value of a specific model might change over time.
- A React-based valuation flow for entering vehicle details
- Integration with a model-backed valuation API
- Interactive views for estimated price, model performance and feature importance
- Market visualizations using charts and geographic data
- An early price-forecasting experience
These screenshots are historical artifacts from the hackathon application and its later marketing iteration.
The primary experience guided users from a vehicle's brand and model into the fuller specification and valuation flow.
The prototype included an exploratory time-series view with a mean-price line and upper and lower bounds. This was a product concept, not a production-validated financial forecast.
| Iteration score view | Error-reduction view |
|---|---|
 |
 |
The preserved interface labelled these charts “accuracy” and “error.” Their exact formulas were not documented in the surviving frontend artifact, so this case study uses R² and MdAPE below as the clearer validation metrics.
The team also explored the geographic distribution of marketplace listings. This map was an analytical prototype rather than a core valuation result.
The project used more than 100,000 real vehicle listings collected from online marketplaces.
Team-reported validation results:
- R²: 0.97
- Median absolute percentage error (MdAPE): approximately 6%
These metrics describe the prototype's validation results; they should not be interpreted as a guarantee for every vehicle or market condition. The forecasting component was an initial exploration rather than a production-validated prediction service.
flowchart LR
User["Vehicle owner, dealer or insurer"] --> Web["React valuation experience"]
Web --> API["Valuation API"]
API --> Model["Machine-learning valuation model"]
Listings["100k+ marketplace listings"] --> Model
Model --> Estimate["Estimated market value"]
Signals["Historical prices and market signals"] --> Forecast["Early forecasting prototype"]
Estimate --> Web
Forecast --> Web
Worthify showed that a small multidisciplinary team could turn a large vehicle-listing dataset and a trained valuation model into an understandable end-to-end product prototype within a hackathon setting.
The original React application, later Webflow marketing export, model implementation, collected data and retired service endpoints are kept private. This repository contains only a concise public case study.
Worthify was a team project. Product names, visual identity, demo media and project materials are shown here for portfolio and historical documentation. See NOTICE.md.