Bengaluru Traffic Police has teamed up with Flipkart and technology partners to explore AI-based ways to tackle traffic congestion in the city. The second
edition of the Gridlock Hackathon drew 33,600 participants from across India and produced more than 1,100 prototypes built around real Bengaluru traffic data.
The challenge was run with MapmyIndia, Arcadis and HackerEarth. Participants worked with anonymised datasets from the Bengaluru Traffic Police ASTraM platform and built models around congestion, traffic violations, movement patterns, parking and deployment of traffic staff.
Dreams don’t always follow an easy path — but determination can take you where you aspire to be.
As reported in The Times of India, 17-year-old J Santhosh, who has 90% spinal muscular disability and cerebral palsy since birth, has secured admission to IIT Madras’ four-year BS… pic.twitter.com/y89tCIqJRL
— IIT Madras (@iitmadras) September 2, 2026
What the winning AI tools try to solve
The first-place project, Gridlock Oracle, focused on predicting how one traffic incident can spread congestion across a larger part of the city. The model is meant to give traffic control rooms an early view so staff can be moved before the situation gets worse.
Second place went to PRAHAR, a predictive traffic system that uses multiple machine-learning models to estimate congestion severity, junction risk, resolution time and possible spillover. It can suggest where officers, barricades and diversion routes may be needed.
The third-place project, ParkSight, studied parking violations to identify areas where illegal or poor parking adds to congestion. It also looked at patrol planning and hotspot forecasting.
Real traffic data used for prototypes
Participants were given access to curated data through ASTraM, a platform designed and implemented by Arcadis for Bengaluru Traffic Police.
The hackathon started with an online machine-learning challenge. Shortlisted teams then moved to prototype development.
The winning teams received ₹2.25 lakh, ₹1.75 lakh and ₹1 lakh respectively.
Rajneesh Kumar, Chief Corporate Affairs Officer at Flipkart Group, said, “Some of the most difficult challenges cities face cannot be solved by one organisation working alone.”
Earlier edition showed measurable gains
The first Gridlock Hackathon was held in 2017 and saw close to 3,000 registrations.
A winning idea from that edition used real-time traffic data for dynamic traffic-signal switching. Testing at Bengaluru’s Silk Board Junction indicated a possible reduction of around 17 per cent in waiting time.
The latest edition takes that idea much further, with AI models now aimed at predicting where congestion may start and how quickly it could spread.
