All work

TAMUhack 2025600 competitors

Kaizen

Vehicle research that weighs what you care about: specs, reviews, and sentiment in one place.

Role

Full-stack developer

Team

Lucas Vadlamudi, Jadon Lee, Sean Hau Goh, Alexander Bui

Links

1st
Place, Toyota Challenge
3
spec and review sources scraped
100+
vehicle features to compare

Stack

  • Next.js
  • Tailwind CSS
  • Flask
  • PostgreSQL
  • Docker
  • Selenium
  • BeautifulSoup
  • OpenAI
  • LangChain
  • Arize Phoenix

Overview

Shopping for a car means tracking more than a hundred features across spec sheets and review sites. Kaizen, named for Toyota's philosophy of continuous improvement, pulls that information together and ranks vehicles by what each shopper actually cares about.

How it works

  • Data. Selenium and BeautifulSoup scrapers collect specifications and reviews from cars.com, toyota.com, and Edmunds into PostgreSQL, running in Docker.
  • Scoring. A weighted scoring system ranks vehicles against each user's stated priorities.
  • Reviews. LangChain and OpenAI summarize consumer reviews with sentiment, so qualitative opinions sit next to the numbers.
  • Trust. Arize Phoenix traces every LLM call, with custom hallucination and correctness evaluations checking the summaries.

Outcome

Kaizen took first place in the Toyota Challenge at TAMUhack 2025. For a four-person team with one shared repository and a weekend, shipping a polished product with few bugs was the accomplishment we were proudest of.

Next project

ThinkTank