ROBERT BORNEMANN

Hyperlocal Predictive Parking · Munich

Parkcast Lab

A prototype that predicts parking availability on a tiny, highly regulated patch of Munich using manually curated data, lightweight ML models, and a phone-style map interface.

Time of day20:00
Likely no parking
Medium chance
Likely some parking
Watch the map cycle through the day. Move the slider to pause and inspect a specific time.

About This Prototype

This is the in-progress Parkcast prototype for hyperlocal parking predictions in my neighborhood in Munich. Parking here is notoriously difficult, so I finally asked the real question: where do I have any chance of finding a spot at a given time? The interface uses a Leaflet map with a time slider and color-coded street segments, each representing the predicted parking availability for that segment.

The current version uses dummy logic for visualization purposes. In the next version, it will connect to the actual ML model and API that processes manually collected parking data.

Key Features

  • Street-Level Parking Predictions
  • Hyperlocal, Rule-Aware Map Overlay
  • Phone-Style Interactive UI
  • End-to-End ML Pipeline

Tech Stack

  • Next.js
  • Scikit-learn
  • React
  • FastAPI