TCA (Toyota Car Agent)

AI. Cars. Toyota. All of which we use and see daily. Car shopping is confusing. With our AI - driven agent, we provide simplification in the process towards finding your new set of wheels.

Comment

Inspiration

We have always been car enthusiasts. The colors, speed, model, and modifications have always fascinated us, to this very day! Now, as adults, we look to buy vehicles for ourselves. After discussing the idea with Toyota representatives, a plan emerged to build this project.

What it does

Our project utilizes a fine-tuned AI model trained on data related to Toyota cars, financing, and leasing to assist users in finding a car that meets their specific needs.

How we built it

First, we used FireCrawl in Python to extract all the data from Toyota's website using web scraping. Then, we created a JSONL file with over 70 entries that contained information about Toyota's cars, as well as details on financing and leasing. Then we fine-tuned the GPT-40-mini model to create ft:gpt-4o-mini-2024-07-18:personal:toyota-advisor:CZqU0cTY. Then we connected to our front-end using Flask API.

Challenges we ran into

Website Scraping - We had trouble with website scraping using Firecrawler. We wanted to filter the data from the Toyota website only based on model. Our program kept getting unnecessary data that was not needed for our agent.

AI Model Connection - We utilized GPT-4 Mini and employed JSONL files to train the model to better comprehend Toyota cars' financing and leasing processes.

Accomplishments that we're proud of

Fine-Tuning an AI Model

What we learned

We learned how to web scrape using Firecrawler, even though none of our teammates had prior experience with it.

What's next for TCA (Toyota Car Agent)

We aim to utilize APIs to connect with online car sellers, enabling users to find listings nearby for the specific car model they are interested in.

Built With

Share this project:

Updates