No hiding behind potholes: Bengaluru engineer’s AI app tracks road damage, names contractors

Bengaluru’s roads remain riddled with deep potholes, many of which have triggered notable accidents and put motorists at risk daily. Turning this every day frustration into a data-driven civic solution, Gaurav Sen, founder of InterviewReady, has built an artificial intelligence system that detects potholes in real time, pinpoints their locations, and identifies the government contractor responsible for repairs.

Real-world test & public response 

During a recent commute to work, Sen’s system detected 12 potholes along his route, compiling documented evidence to lodge complaints with the Bruhat Bengaluru Mahanagara Palike (BBMP).

A video of the setup — shared on Instagram alongside ChatGPT India and later posted on X (formally twitter) — quickly gained traction. Viewers praised the practical tool for tackling a nationwide hazard, urging civic authorities to adopt similar tech to accelerate road maintenance.

How the system works 

Fitted inside Sen’s car, a dashcam, GPS unit, and accelerometer continuously stream video and sensor metrics as he drives. AI vision models process the footage to categorise potholes by size — small, medium, or large — while accurately distinguishing them from standard speed breakers.

Sen leveraged OpenAI’s Codex to develop the underlying application, which captures live data and forwards it to OpenAI models for analysis. This automated detection spots road flaws that are hard to see from a distance, outperforming standard manual inspections.

Linking potholes to govt contracts

Spotting damaged pavement is only part of the process. Sen updated the application to scan roughly 2,900 public contracts, matching each road segment to its specific tender, assigned contractor, and supervising government official.

This link proves crucial when roads remain under warranty, forcing contractors to fix defects without extra taxpayer expense. In just four seconds, the system compiles a complete complaint record containing the image, coordinates, tender ID, and designated officer.