How Netradyne Technology Uses AI In Real Time To Reduce Road Accidents By 50%

Road accidents remain a global scourge, claiming countless lives annually and inflicting a devastating impact on economies worldwide. The fight against road accidents transcends boundaries as governments around the world echo the same question: How can we reduce them?

In 2020, the UN proclaimed the ‘Decade of Action for Road Safety,’ targeting a 50% reduction in road traffic deaths and injuries by 2030. In India, the scenario mirrors this global effort, with Nitin Gadkari, the Minister of Road Transport and Highways, emphasising on road safety and aiming to cut accident deaths by half by 2030.

In the backdrop of this, it may sound uncommon for two techies in the U.S. to dedicate themselves to improving global road safety. But despite the lack of precedence, Stanford alumni and former Qualcomm colleagues Avneesh Agrawal and David Julian embarked on this mission.

They incorporated NetraDyne Technology in 2015 to build tech that analyses driving behaviour in real-time and reduce driver errors and the risk of accidents through automated coaching.

Its vision-based Driver•i safety camera system uses cutting-edge technologies powered by Artificial Intelligence, Machine Learning, and Edge Computing to create a new safe driving standard for commercial vehicles. With inward and outward high-definition cameras on the same IoT device, Driver•i empowers drivers by providing awareness of risky driving behaviour and rewarding safe driving practices.

It analyses 100% of driving time, detecting any driving behaviour that could lead to an accident. This includes over-speeding, driver drowsiness and fatigue, distracted driving, and tailgating, among others, and coaches drivers in real-time. In addition, to incentivise safe driving, the device recognises positive behaviours and rewards drivers with “driver stars” — a gamification element that builds a positive driver profile.

Netradyne is a global company with operations in India, the U.S., Canada, Mexico, the U.K., Australia, New Zealand and Germany. 

From Gaming Box To Deep Algorithms 

“When we started out in 2015, there was no established provider who could combine AI with video telematics to create a solution that could detect, reason, and determine the causality of events. Netradyne stepped in to fill this gap,” said Agrawal.

Netradyne has pooled over 13 Bn miles of proprietary driving data to develop its state-of-the-art analytics models. This massive dataset is the secret sauce behind its machine learning algorithms, which boast a decision accuracy of up to 99%, according to Julian. More importantly, Netradyne has built its AI system from scratch using this unique data.

“Deep tech and innovation are part of Netradyne’s DNA,” Agrawal continued, noting that both founders have over 300 patents combined. According to NASSCOM’s ‘Patenting Trends in India Edition 2024,’ Netradyne is one of the top patent filers among deeptech startups in India.

However, building a tech startup from the ground up was not easy, and the founders faced their fair share of challenges along the way. With a global vision from the outset, Agrawal and Julian constantly connected across different continents and time zones, working tirelessly to solidify their concept. They invested their own capital, rented office space in both India, where Agrawal worked, and the U.S., where Julian worked, and began building their team – most of whom are still with Netradyne today.

Many of the early employees, including office staff, have stayed with Netradyne over the years and have risen to leadership positions. The startup has inculcated a philosophy of innovation, resilience and a dynamic approach since the beginning, which the founders believe also trickles down to employee values.

A good example of this philosophy is how the small team turned adversity into an opportunity during the development of their dataset. The founders quickly realised they needed custom hardware to truly address driver safety. Building their own would take another year or two, Julian explained.

So, they got resourceful. They purchased a set-top box with an NVIDIA processor and used it to run deep learning algorithms in real-time. “We even got a mobile phone and mounted it on the windshield for LTE connectivity, outward and inward cameras, and accelerometer/gyro sensors,” Julian added. (LTE stands for Long-Term Evolution, a standard that improves internet speed for mobile devices).

“To get everything working, our cloud team set up the cloud infrastructure. This allowed us to use a mobile phone for the camera. Since I have a background in deep learning and wireless communication, we sent the data to the GPU on the set-top box for processing and video storage over WI-FI. Finally, WIFI was used to send the processed data back to the mobile phone and then over LTE back to the cloud,” said Julian.

This strategic approach proved successful. It validated the concept and gave the founders the confidence to scale up. With a functional prototype, the founders initiated demos for potential customers and investors, receiving positive feedback. Today, the platform runs a full-fledged cloud repository for data storage, real-time tracking, driver evaluation, and training.

How Netradyne’s AI Wins Driver Trust

The founders believe this system can significantly benefit driver exoneration in cases where the driver is not at fault. Since the camera captures both the driver’s actions and the vehicle’s movement on the road, it provides irrefutable proof in situations where the driver may be unfairly blamed.

“Commercial vehicles tend to be blamed for road accidents. However, more often than not, they are not at fault. This solution addresses that challenge and provides proof,” said Julian.

Netradyne claims that Driver•i has assessed over 13 Bn miles of driving data (driver behaviour, road and traffic patterns) and is clocking more than 600 Mn miles of data every month globally.

Given that the size of data defines the success of any AI system, the availability of such rich data makes Netradyne’s AI highly accurate and precise. The founders emphasised that this is crucial to minimising the number of false alerts for drivers, ultimately gaining their confidence in the technology.

Today, Netradyne maintains a customer base exceeding 3,000 globally, including prominent entities such as IndianOil Skytanking, GreenLine Mobility, Writer Safeguard, CJ Darcl, First Student and Loomis. This extensive network not only solidifies Netradyne’s global footprint but also reinforces its pivotal role in tackling the critical challenge of road safety on a monumental scale. And much like its SaaS peers, Netradyne charges a monthly subscription for its solution, which includes both hardware and software.

Speaking to Inc42, the founders reported revenue of INR 1,000 Cr ARR as of December 2023 and anticipate a 50% year-on-year growth in the current calendar year.

The company has raised $227 Mn across three funding rounds from investors such as Reliance, M12, SoftBank Vision Fund, and Point72 Venture Investments. These funds are being used to drive R&D, enhance core technology, and enter new markets. 

Driving Success With In-House Talent & Tech Expertise

Netradyne is aiming to revolutionise the world of vehicle safety technology by harnessing the power of in-house talent and cutting-edge technological expertise. With sophisticated test labs in India and the U.S., its hardware design and software development teams collaborate seamlessly across borders, with Bengaluru serving as the hub for innovation.

CEO Agrawal recognises the importance of tapping into the best talent worldwide.

“AI talent is very strong in the U.S., but we are also seeing a rapidly growing crop of tech talent in India,” Agrawal explains. This approach allows Netradyne to assemble a diverse team capable of tackling complex challenges.

“An illustrative example of our system’s impact is a case where a driver’s low GreenZone score prompted their spouse to intervene. The spouse’s insistence on safety improvement led to the driver becoming the fleet’s top performer within a month! This highlights the positive influence families can have in encouraging safe driving habits, especially in a profession fraught with danger,” said Agrawal,

For context, Netradyne’s GreenZone is a driver scoring system that assesses individual driving habits and associated behaviour. It is a data-driven approach that provides a quantifiable score by analysing every second of driving. To enable this, the system utilises computer-vision technologies and AI cameras with inbuilt sensors, gyrometer and accelerometers to collect data on driving habits.

The Road Ahead: Connected Vehicle Safety

According to Mordor Intelligence, the global market size for video telematics device installations is projected to surge from 8.3 Mn units in 2024 to 18.8 Mn by 2029, exhibiting a robust CAGR of 17.9% during this period. 

Moreover, the demand for AI-enabled video analytics is steadily escalating, driven by its myriad benefits such as accident prevention, enhanced driver engagement, instantaneous incident notification, and diminished operational costs.

With human error still a major culprit in road accidents, accounting for nearly 94% according to a ResearchGate report, companies like Netradyne offer a crucial solution. By providing real-time feedback, they empower drivers with vital insights beyond their field of view. This real-time coaching fosters better decision-making and a heightened sense of awareness, ultimately leading to safer roads and fewer accidents.

While the penetration of video telematics in trucks stands at around 10-15% in the U.S. and notably lower in India, Agrawal and Julian remain bullish about its rapid uptake in both of these countries as well as other geographies. The founders have already seen a shift, observing that major fleet owners grasp the paramount importance of road and driver safety, and are thus receptive to embracing Netradyne’s solutions.

Having made significant strides in preempting road accidents, it now sets its sights on entering the ‘prediction’ phase to anticipate potential incidents based on the large set of data that analyses the entire road ecosystem on the edge. To this end, it will leverage a foundational driving model utilising large language models (LLMs) based on its proprietary dataset.

Nonetheless, the market presents its share of challenges. Netradyne contends with competition from established players like Lytx and Samsara in California, alongside few India-based providers. However, the founders believe that their innovation: Highly precise and accurate AI algorithms and in-house end-to-end product capabilities set them apart. 

As a startup navigating the convergence of AI-enabled technological disruptions and an expansive market opportunity, Netradyne finds itself positioned at a critical juncture. While the widespread adoption of self-driving vehicles remains years away, the video telematics landscape is poised to flourish, offering a secure passage for human drivers in the interim.