In 2022, India alone witnessed about 4.6 lakh cases of road accidents, in which over 1.6 lakh people lost their lives. While a number of factors contribute to road accidents, one of the major reasons worldover is driver’s negligence. Netradyne, a Bangalore-based, Make-in-India AI solution provider, has been utilising AI and data to address this very problem by providing ADAS (Advanced Driver Assistance Systems) and fleet management platforms. 

“We often see the accident rate reduce by 30% or more after deploying a Netradyne system,” said David Julian, co-founder and CTO of Netradyne. Incepted in 2015, the company focuses on fleet management safety of commercial vehicles. The goal has always been to reduce accidents, especially in the commercial fleet segment. “About 150,000 people die each year due to commercial fleets, and that’s just in vehicle accidents,” he added. 

Data-rich Trained Models

Netradyne’s SOTA models are built from ground up with proprietary data that the company has collected over time. It has a lot of custom aspects and has different approaches for the same. “Turning models into devices at the edge in cost-effective hardware requires a lot of thought on how to actually leverage models for multiple use cases.” 

The decision accuracy falls within a range of 95 to 99%. 

While essentially, there are computer vision models that operate Netradyne’s flagship model Driver.i platform – driver safety system– the scope for generative AI is aplenty. 

“When you look at generative models, what’s interesting to me is that they’re really just trained to do an X token prediction, and with these next token predictions, they actually create all these interesting emergent capabilities,” said Julian. “We look at using our data [both video and sensor data] and do X token prediction to actually create reasoning models.” 

Intelligent Fleet Safety

“If you have a commercial fleet of vehicles, you really want to be able to understand how safe your fleet is. Its intelligence lies in being able to measure the safety of your fleet, understanding who your safe drivers are and who are the ones driving in an unsafe manner – and that is intelligent fleet safety,” said Julian. 

The company has analysed over 10 billion driving miles and claims to have brought a 96% reduction in distracted driving. It has created a toolset that fleet managers can deploy to gain visibility into their fleets and coach their drivers to improve safety. 

Coaching happens in various forms, including one where a fleet manager guides and coaches the driver based on Driver-i systems, and real-time coaching feedback, where the system alerts the driver before getting into accidents. “We even get into predictive mode where it can say, hey, looks like you are going to fall asleep in the next 15, 30 minutes or something like that.” 

Trained For Demographics

Netradyne’s initial two markets were US and India. “Driving is handled in these two environments very differently. The algorithms have been trained from the ground up and from early on to really be able to tackle these different types of environments,” said Julian. 

“We came from a structured driving environment to a dynamic driving environment, and developed our AI and algorithms to be able to operate in any country in the world.” 

However, demographics also pose various challenges when it comes to training the models. Bringing solutions to various markets mandates the understanding of different driving laws clubbed with the different make of vehicles. “How do you make the driving environment work across areas?” is the focus for global expansion. Apart from the US and India, Netradyne products are used in Australia, New Zealand, UK, Germany, Canada and Mexico. 

With the existing mechanisms, Julian believes that giving a tech upgrade will help improve the model. “I think there’s a lot of opportunity for additional cameras and things like that so that you can actually start to understand the pedestrian intent – is the pedestrian about to walk in front of you, and so forth.” 

No matter how different the ecosystems are, there are commonalities too. “I think one of the things we actually are able to leverage is that there are a core set of features that tend to be common across localities. So, when we enter a new environment, we can enter with these core features,” said Julian. Over time, this data can be tuned based on the audience. 

Growing Indian Market

Netradyne has lately entered into partnerships with crucial players in the Indian market. The company recently partnered with GreenLine Mobility Solutions, which operates a fleet of liquefied natural gas (LNG)-powered trucks in India, to provide their flagship Driver.i system. 

Furthermore, a few weeks ago, Netradyne announced its collaboration with IndianOil Skytanking, India’s foremost aviation fuel management and airline fueling service provider. The company currently operates in the commercial segment, working with long-haul trucks and gas-delivery vehicles, and the company has plans to provide solutions for passenger vehicles in the future. 

“I think one of the key things for commercial vehicles is that they tend to drive much more than individual passenger vehicles by offering an order of 5 to 10x. Therefore, the economics worked out much better in the long term, that we can actually see our tech potentially into personal vehicles, for smart insurance and other things as well as driving in support,” said Julian. 

“Our global vision is to get into every vehicle out there and really make a difference in driver’s safety. At the end of the day, get people home safely.” 

The post Bengaluru AI Startup Netradyne Cuts Road Accidents by 30% appeared first on Analytics India Magazine.