Bengaluru techie turns car dashcam into AI tool that spots potholes and files a complaint in four seconds with photo and exact location

A Bengaluru engineer has used a car dashcam, GPS and AI to tackle a problem that has become part of everyday driving in the city: potholes. His system not only detects damaged roads but also records their location and prepares a complaint in seconds, while trying to identify who is responsible for fixing them.

Also read: El Niño effect: Bengaluru bans drinking water for car washing, gardening amid monsoon disruption

One bumpy commute led to an AI experiment

For Gaurav Sen, the idea started with a frustrating drive through Bengaluru. He encountered 12 potholes during one commute and decided to see if technology could make reporting them easier.


Sen fitted his car with a dashcam, GPS and an accelerometer. The setup records the road as he drives, along with location and movement data that can help identify where a pothole was encountered.

He then used Codex to build an app that processes the information and sends it to AI vision models.

The result is a system that can scan the recorded footage, spot potholes and classify them by size, including small, medium and large defects.

Sen shared the project on Instagram on August 5, describing it as his way of reporting potholes to the Brihanmumbai Municipal Corporation, or BBMP. He also pointed to the wider danger posed by damaged roads, saying more than 2,000 Indians die every year in accidents described as pothole-related.

The app does more than spot a pothole

Detecting a pothole is only the first step.

A photograph of damaged road may show the problem, but a complaint still needs details such as where the pothole is located and which authority or contractor is responsible for that particular stretch of road.

Sen's system attempts to solve that part too.

The app searches through around 2,900 government road contracts to connect a detected pothole with the relevant road project. It can identify the tender number, the contractor associated with the work and the government office responsible for the road.

That could be particularly useful for roads that are still covered by a maintenance or construction warranty. If a road is under warranty, the contractor may be responsible for repairing defects rather than the repair being treated as a fresh government expense.

View this post on Instagram

A post shared by Gaurav Sen (@gkcs__)

A pothole can become a complaint in four seconds

The most striking part of the system is what happens after the pothole has been detected.

Instead of leaving the user with a photograph and coordinates, the app puts the information together into a complaint record. It can include the pothole image, exact location, tender details and the relevant government office.

Sen says the system can prepare this information in about four seconds.

In other words, the process moves from spotting a pothole to having the basic material needed to report it without requiring the driver to manually record every detail.

One Bengaluru drive found 12 potholes

Sen tested the system during his regular commute and the app detected 12 potholes along the route.

Each detection could be tied to photographic and location information, creating a record of the road damage rather than simply relying on a driver's description of where the pothole was.

The system also attempts to take the complaint one step further by connecting the road damage to the contract and authority responsible for that stretch.

That is important because identifying a pothole and getting it repaired are two different problems. A civic complaint can only go so far if the person reporting the problem has no idea which agency or contractor is responsible.

Why the idea has attracted attention

The project has drawn attention because it applies AI to a problem that Bengaluru motorists encounter every day.

The technology itself is relatively simple in terms of hardware. A dashcam captures the road, GPS provides the location and an accelerometer records vehicle movement. AI then processes the footage and helps identify the damaged sections.

The more unusual part is the attempt to connect that road damage to government contracts.

Instead of treating a pothole as an isolated defect, Sen's system tries to create a chain linking the pothole, photograph, location, road contract, contractor and responsible office.

That turns an ordinary commute into a source of documented civic information.

From detecting potholes to making someone responsible

Sen's experiment is ultimately less about building another pothole detector and more about making the reporting process easier.

A driver can spot a pothole, but documenting it, finding the right authority and preparing a complaint can take considerably more effort. His system attempts to compress those steps into a few seconds.

The idea also raises a larger possibility for civic technology. If similar systems could collect road damage data at scale, cities could potentially build a much more detailed picture of where roads are deteriorating and which projects or contractors are linked to those stretches.

For now, Sen's project remains an individual experiment built around his own commute. But the basic idea is straightforward: spot the pothole, record the evidence, find who is responsible and make the complaint ready to file.