ParkPulse
Enforcement Intelligence
Gridlock Hackathon 2.0 · Submission

The 3-minute pitch

The full ParkPulse story — the problem, the pipeline, and a live walkthrough of the dashboard you're looking at. Watch the video, then explore the tabs in the sidebar.

Trouble with the embed? Open it directly on YouTube.

Watch on YouTube

What to watch for

ParkPulse runs one spine end to end — each step maps to a tab in the sidebar.

Detect

298,000 challans collapse into a live hotspot map — where violations actually cluster, not where we guess.

Score

Every zone gets a Congestion Impact Score, so a blocked arterial outranks a quiet lane with the same ticket count.

Forecast

A Bayesian-shrunk model predicts demand by weekday × hour — validated on held-out weeks, not inflated in-sample.

Deploy

Greedy max-coverage allocation places patrol teams 600 m apart for the chosen window — no overlap, no blind spots.

Target

Repeat-offender vehicles surface with a recommended action — owner notice, escalated penalty, or tow priority.

Ask ParkPulse

A Gemini co-pilot turns plain English — “plan six teams for Friday evening near KR Market” — into a deployable plan, and is honest when a window is a blind spot.