CarBaba

Inspiration

The inspiration for our project came from the lack of a good website that provides a definitive experience of comparing different cars and helping the user truly decide which option fits their needs the best. A lot of websites try to sell fancy names to customers without truly showing them all the options they have. The pain of having to scroll through endless UIs and ask countless people for opinions is what compelled us to come up with a solution.

What it does

Carbaba is an AI-powered car search and recommendation system. It fills in the gaps that aren't addressed by the website options available today, by providing personalisation on a granular level. Not only does it allow users to search for cars with the exact features they are looking for, but it also adds a layer of customizability by allowing users to set weights for their preferences. It provides a truly immersive experience with an AI bot that understands natural language and uses an advanced search algorithm to find the correct matches for the user.

How we built it

The application was built using React.js + Vite on the Frontend, and the backend was orchestrated in Python with FastAPI. The Gemini API powers the AI bot, and ElevenLabs powers the speech-to-text and text-to-speech features. The database was stored inside MongoDB Atlas. Deployment for the backend was done on a VM on the Google Cloud Platform. The frontend deployment was done on Vercel.

Challenges we ran into

Some challenges we ran into included finding 3D models of the cars to elevate the buyer experience. We were able to scrape some models from an online website. Finding data was also initially difficult, so we had to scrape from multiple websites. Coming up with the recommendation system was a complex task, which needed a lot of trial and error

Accomplishments that we're proud of

Some accomplishments that we are proud of include the usability of our recommendation system, providing really granular control over preferences. Our UI is smooth with clean animations and transitions, and the 3D models elevate user experience. We also provide a direct comparison mode in Carbaba to compare any two cars on all parameters.

What we learned

We learnt a lot about webscraping and building web apps. Building our own recommendation algorithm gave us insight into how recommendation systems work.