🚀 About the Project – VelocityAI 🏎️⚡
🏁 Inspiration
Formula 1 is a sport driven by speed, strategy, and data. As a passionate F1 fan, I wanted to create an AI-powered chatbot that delivers real-time race insights, driver stats, and historical analysis at lightning speed. My goal was to make F1 knowledge easily accessible to fans, analysts, and enthusiasts alike.
🛠️ How I Built It
VelocityAI is powered by OpenAI, LangChain, and FastAPI on the backend, with DataStax AstraDB as a vector database and Next.js handling the frontend. I integrated:
- Web Scraping (using Pyppeteer) to extract F1-related data from trusted sources.
- Vector Search (using DataStax AstraDB) for efficient information retrieval.
- Retrieval-Augmented Generation (RAG) to enhance AI responses with real-world F1 data.
- Custom Embeddings to store and retrieve relevant F1 knowledge efficiently.
📚 What I Learned
- How to efficiently scrape, process, and embed data for AI-powered applications.
- The power of vector databases in storing and retrieving structured information.
- Optimizing retrieval-augmented generation (RAG) to improve chatbot accuracy.
- Managing rate limits and API constraints when working with large-scale AI models.
🚧 Challenges I Faced
- Extracting structured F1 data from dynamically loaded web pages.
- Fine-tuning embeddings for accurate question-answer retrieval.
- Handling API rate limits while processing large datasets.
- Structuring the backend for scalability and efficiency.
🎯 Future Improvements
- Adding live race updates for real-time analysis.
- Expanding knowledge with historical F1 race archives.
- Integrating text-to-speech for an immersive voice-based experience.
VelocityAI is built for speed, precision, and intelligence—just like an F1 car! 🏎️💨
Built With
- astradb
- datastax
- fastapi
- langchain
- nextjs
- openai
- pyppeteer
- versal
Log in or sign up for Devpost to join the conversation.