Bengaluru engineer builds AI system that detects potholes and prepares complaints

Driving through Bengaluru roads regularly, potholes are likely an all-too-familiar part of your daily commute. They can damage vehicles, make roads unsafe and leave you wondering who is actually responsible for fixing them.

One Bengaluru-based engineer has now built an AI-powered system that attempts to answer that question and turn a road defect into an actionable civic complaint.

How the AI pothole detector works

Gaurav Sen, founder of InterviewReady, shared a post on X about how the AI pothole detector. He developed the system after getting tired of potholes damaging his car. Built with Codex, the setup uses a dashcam fitted with GPS and an accelerometer to record the road as he drives.

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

The system uses an AI vision model to detect potholes and classify them based on size - small, medium or large. It also attempts to distinguish potholes from other road features such as speed breakers.

That means you don't have to stop every time you spot a damaged stretch of road and manually record its location. The system can capture the information while you continue driving.

It goes beyond simply spotting potholes

The more unusual part comes after a pothole is detected.

According to the demonstration, the system searches through roughly 2,900 government contracts to identify the tender linked to the road. It then finds the contractor responsible for that work and identifies the relevant government officer.

Must Read: Anthropic CEO Dario Amodei predicts AI could eliminate most human illnesses in next 5 to 10 years

This matters because road contracts can include maintenance and defect-liability provisions. In such cases, contractors can be responsible for addressing defects during the applicable period rather than leaving citizens to simply report the problem.

The app then attaches the pothole photograph and its geographical coordinates before generating a structured, ready-to-file complaint.

One drive, 12 potholes and 12 complaints

The demonstration showed how quickly the process could work in practice. During a single drive to work, the system detected 12 potholes and generated 12 complaints ready to be filed.

Bengaluru already has civic mechanisms for reporting potholes, including the former BBMP's geo-tagged “Fix My Street” system and the Greater Bengaluru Authority's “Fix Pothole” service. The AI system adds another layer by attempting to connect each road defect with the contract, contractor and official responsible.

For you as a commuter, the bigger idea is simple, instead of AI merely telling you where a pothole is, it could help identify who needs to act on it.