Independent scorecards and retiming plans for city traffic signals
TL;DR Gridtune reads the cameras and controller logs a city already owns and reports, junction by junction, whether its signals are working, adapting and timed well. It then hands engineers revised timing plans. Traffic police and smart city companies pay a yearly fee per junction for proof that their signal spend is paying off.
Indian cities are spending on smart signals, but nobody outside the project checks the results. Bengaluru now runs 169 adaptive junctions[1]. Its travel-time claim of up to 25% comes from a case study written by one of the system integrators, with no baseline or method given[1]. Police statements at launch spoke of "significant reductions" without a single figure[2].
Conditions on the ground make drift likely. A vendor in the space notes that many Indian junctions still run old fixed-time controllers that can't take live timing updates, and that roadside power and network links fail[3]. When a camera dies, an adaptive junction quietly falls back to a fixed plan and nobody gets told. How often does that happen? I found no public audit that says.
What it does, in regulatory terms: analytics and advice only. Gridtune never switches a signal itself. Engineers at the city decide and load any new plan.
It plugs into the junction cameras and controller logs a city already has, counts vehicles by type and approach, and measures what actually happens - phase lengths, queue spillback, stops along a corridor and how many hours each junction ran in adaptive mode versus fallback. Every month the city gets a scorecard that ranks junctions by delay and by faults. For the worst fifty it proposes retimed fixed plans or offset changes, then measures again after the change.
The seed was a startup selling its own real-time controller, and I've dropped that. It would have to compete with C-DAC's CoSiCoSt software installed through large integrators[2] and take on safety liability for every red light. A neutral measurement layer works alongside whichever controller is installed, and that neutrality is what gives its numbers weight.
Drivers benefit, but the buyer is a traffic police unit, a municipal corporation or a smart city company that has already paid an integrator and now needs proof for councils and auditors. Pricing is a yearly fee per junction plus paid retiming studies. There's no public price for comparable services in India, so test it. Integrators are a second buyer, since independent figures help them defend their own results.
Bengaluru's system moved into full operations in April 2025 with 400 or more junctions on its roadmap[1], and other cities will copy it. Camera networks are already installed for enforcement, so the data exists[3]. Central buyers must use GeM where a service is listed[4], which gives a small vendor a route that doesn't depend on a large tender.
There's no public national count of adaptive or signalised junctions in India. The narrowest firm figure is Bengaluru, with 169 adaptive junctions live and about 400 more planned[1][2]. Adjacent - SCATS, the Australian system, runs at more than 63,000 intersections in 216 cities worldwide[5], which shows how far signal control can spread once cities commit.
C-DAC with a global engineering firm runs Bengaluru's system and reports its own results[1]. A Coimbatore start-up sells camera-based adaptive control and traffic management[3]. A US start-up sells AI sensors with a cloud dashboard and a 24/7 operations centre, and operates in 40 states[6]. All of them sell the control system. Nobody in India sells independent, controller-neutral measurement, and that's the opening.
| Difficulty | Moderate (49/100) |
| Starting capital | ₹1.2–2.2 Cr |
| Time to launch | 9–15 months |
The vision models aren't the hard part. Getting read access to camera feeds and controller logs owned by an integrator is. Each city has different hardware and log formats are often closed, and government sales cycles are long, with payment lagging delivery by months.
No product licence, because Gridtune advises and never operates a signal. The real gates are procurement and data access.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI push timings to a live signal. Engineers must decide and load any plan, because a wrong red light is a safety issue, not a bug.
Want to work out the AI layer for Gridtune? Talk to Yash →
I'd back this for a founder who has worked inside a traffic police unit, an integrator or a transport consultancy and knows how city signal contracts work. A pure computer-vision team will stall at the data-access stage, no matter how good the models are. Prove one thing first - that one city will give read access to 30 junctions and act on a retiming plan.
Plan for roughly ₹1.2–2.2 Cr, since data integration across cities and slow government payments take time and money.
No product licence. You need GeM or tender registration, written data access from the city and DPDP compliance.
Yes, cities everywhere buy adaptive signals and need proof they work. Check local procurement rules and who owns the controller data.
The free Gridtune validation kit has the line-by-line capital breakdown, the effort scoring, the procurement and data-protection checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for traffic police and integrators, and 5 assumptions to test. Download the kit →
Thinking about building Gridtune or something close to it? The make-or-break work is winning data access from the integrator and a first paying city, and that's the kind of thing I help founders with. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.