Revvo

Rating & insurance calculations and car info
AI chatbot assistant
Profile settings that factor into our car listing recommendations
Car listings and filters
Landing page, sign in and user authentication

Inspiration

Buying a car is one of the biggest financial decisions people make, but the process is often unclear and overwhelming. Traditional car buying platforms show car listings but don't help buyers understand the true costs of ownership, hidden issues, or whether they're getting a good deal. We wanted to create "the world's most honest car dealer," a platform that assists buyers with AI-powered insights, transparent pricing predictions, and honest LLM conversations about every vehicle.

What it does

Revvo is an intelligent car buying platform that transforms how people shop for vehicles. Users start by setting up a profile with their budget, location, and car preferences. Our ML model then generates personalized car recommendations and fetches real listings from Auto.dev. For each car, we provide:

Smart Car Insights: Maintenance history, reliability scores, and known issues of that make and model
AI Chat Assistant: An AI chatbot that answers questions about specific cars, evaluates deals, and provides expert advice
Insurance Cost Estimation: Detailed breakdowns showing how location, make, age, mileage, and accident history affect insurance costs
Depreciation Graph: Interactive chart showing how the car's value will change over time
Comprehensive Ratings: Ratings for the quality of the deal, fuel economy, maintenance, safety, and owner satisfaction
Vehicle History: Accident counts, ownership history, and car usage type information

How we built it

Frontend: Built with React, TypeScript, and Vite for fast development. We used Framer Motion for smooth animations, Recharts for data visualization, and Firebase for authentication and user profile storage. UI has a gradient design with slow movement.

Backend: Flask-based Python server with modular architecture. We integrated OpenAI's GPT-4o-mini for:

Generating personalized car recommendations based on users’ preferences
Powering conversational AI chat about specific vehicles
Generating car ratings

APIs & Data:

Auto.dev API for fetching real car listings with vehicle details, images, and dealer information
Custom insurance prediction model using weighted averages for state, make, body style, engine size, age, mileage, accident history, and ownership patterns
Dual-exponential depreciation model for forecasting the vehicle’s value over time

Deployment: Configured for Vercel with serverless functions for API endpoints, which enables scalable deployment.

Challenges we ran into

API Integration Complexity: Coordinating multiple APIs (OpenAI, Auto.dev, Firebase) with different response formats and rate limits required careful error handling and data transformation
Data Parsing: Auto.dev returns complex nested JSON structures that needed cleaning and normalization to obtain usable vehicle information
Insurance Model Accuracy: Building a heuristic-based insurance prediction model that accounts for multiple factors (location, make, history, etc.) while maintaining reasonable accuracy without access to proprietary insurance data
AI Response Parsing: Ensuring OpenAI responses are consistently formatted as JSON for recommendations and ratings, handling edge cases where the model returns text instead of structured data
Real-time Chat State Management: Managing conversation history and context in the AI chat feature while maintaining good UX with loading states and error handling
Depreciation Calculations: Implementing a depreciation graph that accurately reflects real vehicle value curves

Accomplishments that we're proud of

Insurance Breakdown: Created an educational insurance estimation that shows users exactly how different factors (location, make, accident history, etc.) impact their insurance costs, with visual breakdowns
AI-Powered Personalization: Successfully integrated OpenAI to provide intelligent, context-aware car recommendations and conversational assistance
UI: Built a clean interface with smooth animations, and interactive images
Real-Time Data Integration: Successfully aggregated real car listings from Auto.dev with AI-generated insights, ratings, and predictions
Full-Stack Architecture: Delivered a complete full-stack application with user authentication, profile management, and data persistence with Firebase

What we learned

API Use: Learned how to effectively chain multiple APIs together, how to handle error cases and data transformation
AI Integration: Gained experience with prompt engineering, JSON parsing, and managing AI responses
Data Visualization: Explored creating charts and graphs (depreciation curves, insurance breakdowns)
React: We implemented hooks, context API, and component composition patterns for more maintainable code
Serverless Deployment: Practiced using Vercel's serverless function architecture and how to structure projects for scalable cloud deployment

What's next for Revvo

More nuanced AI Recommendations: Expand the recommendation engine to consider more factors like lifestyle, family size, and long-term ownership goals
- Price Negotiation Tools: Add features that suggest the best negotiation strategies and fair price ranges based on market data
- User Reviews: Incorporate verified owner reviews and ratings to on top of our AI-generated insights
- Saved Searches and Alerts: Allow users to save favorite cars and set up alerts for price drops or new listings matching their criteria
- Finance Calculator: Integrate loan and financing calculators to show total cost of ownership including interest rates
- Mobile App: Develop mobile applications for iOS and Android users
- Dealer Integration: Partner with dealerships for direct communication and appointment scheduling through the platform Machine Learning Improvements: Train custom ML models on historical car data to improve insurance predictions and depreciation forecasts