Inspiration
Every Lehigh student knows the struggle of getting a parking ticket. We wanted to make a tool that helps students avoid tickets before they happen by showing real-time data on where they’re being issued.
What it does
Parq is a web app that alerts you to the latest parking tickets issued at Lehigh University and shows how far each one is from your current location.
How we built it
- We scraped public parking citation data and cleaned it into a unified format using Python. The data is stored in a Supabase database and pushed to an AWS S3 bucket for model training.
- Using AWS SageMaker, we trained a first-order Markov model to predict the next likely ticket location. The trained model is saved to S3 and served through an AWS Lambda function connected to an API Gateway, which powers the AWS Amplify-hosted frontend. The map visualization uses GeoJSON and React to display tickets dynamically ## Challenges we ran into Getting real-time data from the parking website without breaking rate limits. Deploying Python packages like h5py and numpy on AWS Lambda, which required custom layers. Cleaning inconsistent address data and mapping it accurately to coordinates. Integrating multiple AWS services smoothly under tight hackathon time limits. ## Accomplishments that we're proud of Built a fully cloud-deployed ML pipeline from data scraping, model training, and live prediction ## What we learned We learned how to connect machine learning workflows with cloud infrastructure, including SageMaker, Lambda, and Amplify. We also gained hands-on experience with data cleaning, encoding geographic information, and managing serverless deployments. ## What's next for parq Add push notifications when tickets appear near you. Improve prediction accuracy with spatial clustering and time-based analysis. Expand beyond Lehigh to other universities or cities
Log in or sign up for Devpost to join the conversation.