Inspiration
The need for a tool that quickly looks at cars and then gives the vehicle's expected value and conveys and visible concerns.
What it does
Our program uses a large CSV file from craigslist's historical data, then runs linear regressions on the the data depending on the initial information provided. The program then passes images of the car to the Gemini API, and returns a multiplier on price based on the condition of the car.
How we built it
We built this program using python streamlit to do the user interface, and plain old python to do the entire backend logic.
Challenges we ran into
The Gemini API stopped working at the end, we hit the token limit. We were also having a problem with the statistics library which kept saying missing variable.
Accomplishments that we're proud of
The API was integrated correctly, and the LLM performed its task correctly. This was an accomplishment because none of us had ever done this before. Cleaning up the the CSV file was also an accomplishment as we had to use python scripts to throw away data that was not relevant to us. The original CSV file was 1.3 Gb.
What we learned
We learnt how to integrate API's into our projects, how they work, and how to exactly use them for reasoning. Another thing we learnt was how to clean up data using python, and then reading files effectively even when the amount of data is massive.
What's next for AutoCurve Assistant
We need to find a way to make the Gemini API work properly, as it hit its limit and told us to wait quite quickly just when testing out program. We plan to finish this project soon. Finishing this program can help real people quickly judge the values of real vehicles.
Log in or sign up for Devpost to join the conversation.