For Bengaluru engineer Gaurav Sen, potholes became more than a source of bumpy rides and vehicle damage.
He built a system using a dashcam, AI and government contract data to detect potholes and identify the contractors responsible for maintaining the roads.
Sen demonstrated the system in a video shared on Instagram in collaboration with ChatGPT India. The video later circulated on X and attracted attention for applying AI to a common civic problem.
The system uses a dashcam, GPS and an accelerometer fitted in Sen’s car to collect data while he drives. AI vision models then analyse the footage and identify potholes of different sizes. Sen said he used Codex to build an app that records the data and sends it to OpenAI.
The system can also distinguish potholes from features such as speed breakers, which can appear similar from a distance.
Sen then modified the app to search through 2,900 government contracts and identify the contractor responsible for a particular stretch of road. He said this could be useful when roads are still under warranty, and contractors are required to repair defects without additional cost to the government.
The app can generate a structured record within four seconds, containing the pothole’s photograph, exact location, tender number and the officer responsible for the road.
Sen tested the system during a drive to work, where it detected 12 potholes along his route. The resulting records could provide documented evidence when complaints are filed with civic authorities.
After the demonstration circulated on X, users praised the system and suggested that civic authorities could use similar technology to detect damaged roads and accelerate repairs.