How India-built AI is transforming global road safety

Time of India

5 days ago

A safety revolution is being scripted inside fleet vehicles — not with seat belts or speed governors, but with high-powered cameras, edge processors and AI models trained to detect everything from drowsy drivers to heavy snow.
Tired of too many ads? go ad free now
One of the leaders in such technology is Netradyne, which has its R&D engine in Bengaluru and a footprint that now spans 10 countries – soon to be 15. Backed by Reliance, SoftBank, Microsoft's M12 and Qualcomm,
Netradyne
recently joined India's unicorn club after raising $90 million in Series D funding. Its core product is an advanced
computer vision system
that analyses real-time driving behavior via dual/ quad HD cameras and edge AI – delivering live audio alerts to drivers and analytics to fleet operators operating 450,000 vehicles worldwide.
It detects risks such as drowsiness, unsafe following distance, or traffic violations, and issues instant in-cab alerts. The system also captures positive driving actions, generating a composite driver score that correlates with accident risk.
'We began in the US and India deliberately because driving conditions are polar opposites. If our models could handle both, we'd have a globally adaptable baseline,' Durgadutt Nedungadi, senior VP for EMEA and Apac at Netradyne, says.
Early tests with off-the-shelf hardware, including SD cards and SIM slots, failed under temperature and power constraints, prompting the company to design a custom edge platform optimised for thermal conditions, latency and realtime compute.
David Julian, CTO and who co-founded the company with Avneesh Agrawal (both were in Qualcomm for many years before founding Netradyne in 2015), notes that edgefirst processing, instead of cloud-based decisions, remains central to the platform.
Tired of too many ads? go ad free now
'When you're delivering incab alerts, you need a very high precision rate. Otherwise, the driver loses trust,' he says. The company has built out a feedback loop that moves data from the field to labelling pipelines and back to the device via over-the-air updates, allowing it to continuously improve model accuracy in live environments.
The platform's real-time coaching capability was a major evolution from earlier generations of driver safety systems.
'Legacy systems relied on days-later video reviews. We wanted to cut that feedback loop to minutes, even seconds,' Julian says. 'That shift led to measurable improvements in compliance and safety.'
Netradyne's use cases span last-mile delivery vehicles, school buses, oil and fuel tankers, and employee transport fleets. In India, its clients include Shell, Indian Oil Skytanking, Writer Safeguard, and Greenline.
'We're also seeing strong adoption in tier-2 cities where last-mile logistics is growing rapidly,' says Nedungadi.
The tech platform has evolved to support facial recognition, seatbelt compliance, U-turn detection, and predictive warnings such as forward collision and pedestrian alerts. Netradyne is also among the few global players capable of traffic light analytics on the edge, a technically complex feature that factors in signal states, lane positioning and surrounding traffic to assess driver behaviour.
TRACKING POSITIVE BEHAVIOUR
Its unique ₹driver stars' functionality recognises positive behaviour – for instance, if a driver proactively slows down to allow safer merges. 'That's actually harder to detect than violations, because it requires deeper contextual understanding,' Julian says.
The platform processes up to a billion miles of driving data monthly and generates a ₹GreenZone' score, akin to a credit score, which aggregates risk and compliance factors.
This feeds into a virtual coaching system that prioritises the most critical behavioural corrections for each driver, pushing targeted video training and app-based feedback loops.
In colder regions, Netradyne has added modules for snow and fog detection, now live in parts of the US and Europe.