How Edge AI Is Transforming Driver Safety | Ep 18 | Driving Innovation by Sonatus

In this episode of the Driving Innovation podcast, Siva Yoganathan, VP of Engineering at MOTER Technologies, discusses how edge AI and foundational models are transforming automotive insurance into real-time, contextual driving intelligence. He also highlights MOTER's partnership with Sonatus to streamline model deployment across diverse OEM platforms while improving privacy, processing speed, and cost efficiency.

Creators and Guests

Host

Sanjay Khatri

Sanjay Khatri is Head of Product Marketing at Sonatus, Inc., a leading provider of software solutions accelerating the shift to AI-enabled Software-Defined Vehicles (SDVs). He brings more than 20 years of experience in product management and marketing for cloud and edge software across Fortune 500 companies and early-stage startups. At Sonatus, Sanjay defines product positioning and go-to-market strategy for solutions that help automakers deliver intelligent, adaptive vehicles that evolve with changing market demands. His background includes leading strategy and execution across Connected Car, In-Vehicle Infotainment, and IoT technologies. Sanjay holds a Bachelor’s degree in Computer Engineering and an MBA, along with certifications in AI/ML, telecommunications engineering, and content strategy. He is a frequent writer and speaker on the future of automotive software, edge AI, and the rapidly transforming mobility ecosystem.

What is Driving Innovation by Sonatus?

Dig into wide-ranging technical topics about modern vehicle technology with industry leaders, hosted by Sanjay Khatri, Head of Product Marketing, Sonatus.

Welcome to Driving Innovation,
a podcast from Sonatus that

explores technologies and
solutions that are defining the

automotive and
mobility industries.

Today, we're exploring how AI at
the Edge is transforming driver

risk assessment,
insurance innovation,

and in vehicle intelligence.

Joining me is Siva Yoganathan,

Vice President of Engineering
at MOTER Technologies.

MOTER is helping redefine
Insurtech by moving beyond

traditional underwriting models
and leveraging real world

driving behavior,
contextual vehicle data,

and embedded AI to create safer,

more personalized
mobility experiences.

In this conversation,

we'll discuss how MOTER and
Sonatus are working together to

deploy advanced AI models
directly inside the vehicle

using Sonatus AI Director,
enabling real time driver

coaching, scalable deployment,
stronger privacy protections,

and a fundamentally new
approach to behavior based

insurance and safety.

Sival, welcome to
Driving Innovation.

Let's start with a quick
introduction to and your role there.

Thank you, Sanjay, for
having me on this podcast.

So MOTER is an insurtech
company and we are focused

on developing AI foundational
models that are used

for risk intelligence
analysis on the edge.

Our focus is not just
about measuring risk,

but actually about
reducing risk.

And also about not really
using the vehicles for data

collection, but actually
building AI models that would

help with interpreting
them, analysing them,

and bringing value
right on the edge.

And that's why MOTER
focuses and as the name

suggests, MOTER is "Mobility
on the Edge in Real

Time".

Okay.

So that's our name.
That's clever.

Thank you.

So as I understand it,

MOTER's approach goes well
beyond traditional usage based

insurance models.

Can you explain what's
fundamentally different about

your driver risk analysis model and
why contextual driving intelligence,

you put a lot of emphasis on that
matters so much in this context.

Right.

So when you think about
a traditional UBI,

they're mostly using brake
acceleration turns based

events, right, and speeding.

It's true they bring value.

But it's important to
understand the context in which

particular driver may
be applying a brake.

Let's take an example of
a hard braking, right?

So a person may be applying a
hard brake because an object

came in front of the
vehicle in the same lane.

So that is avoidance.
That's wonderful.

But what if the braking is
due to a prolonged aggressive

tailgating or a lane distraction
or excessive lane departure?

So those are risky events
that can be corrected

and can lead to less
collisions, right?

Therefore, contextualisation
is important.

So in our foundational
model building,

we look at events by looking
at the road scene inside the

cabin and the basic telematic
information to bring

a collection of AI foundational
models that are approved

by Department of Insurance
for assessing risk.

Great, so you want to
be able to eliminate false positives.

If somebody is braking
for legitimate reasons,

you don't want to
penalize them for that.

That totally makes sense.

One of the most compelling
aspects of this collaboration

is that these models run
directly inside the vehicle.

Why was edge execution
so important for MOTER?

Right.

Sanjay, I've been in the
mobility space for a long time

and I've worked on IoT
and V2X and now I'm at

insuretech with
MOTER, where again,

it's so important that we are able to
provide feedback in near real time.

Take an example today.

On a typical day in
the United States,

tens of millions of
vehicles are on the road.

Imagine if each vehicle
has to send all the vehicle

data to the cloud to
be processed, right?

And then some of
that key information,

the consumer is in the vehicle.

You've got to send it back
to the driver to be consumed.

So it's not effective, right?

So when we are able to apply
these AI foundational models on

the edge, these phenomenal use cases
can be consumed and used immediately.

So that's why an opportunity
for us to be able to

run our models on the edge
is very important, right?

It really streamlines
the process.

Large data transmission
is cut off,

your user experience
in the vehicle,

and privacy is a
key thing, right?

Why do we have to send all the
data to the cloud when it can

be processed and analyzed
right on the edge?

Let's talk about Sonatus
AI Director specifically.

What role did it play in making
this deployment production ready?

So from an insurtech company
like ours perspective, right,

our core competency is
driver behavioural analytics.

So we spend a considerable
amount of time building these

complex AI foundation models.

But that's just one part of
the problem. How do we deploy that?

You think about it, there
are several dozen OEMs.

Each has many makes and
models. Yearly they distribute.

And underneath that,
there is several hardware

infrastructure, OS
infrastructures.

How do we then deploy
these models? Right?

So your AI Director
is an amazing

example, a great solution,
where it actually abstracts all

that complexity
away from us, right?

We focus on our core competency of
driver behavioral risk analytics.

We build those models. We apply
them in your infrastructure.

Once we have an OEM
relationship, yes,

that's necessary.

But the implementation
infrastructure, you take care of it.

Right? You abstract us from
all those complexities.

And therefore, it makes it so easy
for us to be able to deploy across

various OEMs, different
brands, and still provide the

same experience to the users.

Yeah, it makes a lot of sense.

Now, your platform produces
both insurance scoring outputs

and driver coaching insights.

How do you see OEMs and
insurers benefiting from those

dual capabilities?

You know, one of the
challenges today in insurance,

in the safety area, is
the fragmentation, right?

So

insurance likes to measure risk
of the particular driver and

also the entire
portfolio, right?

And then the driver is focused
about what is it they can

improve on so the safety
improves and also potential

reduction in premium.

So you have to have a single
collection of foundation AI

models that every entity,
every stakeholder in that

ecosystem, mobility
ecosystem is using, right?

OEMs, right? So they
also have a need, right?

So they're spending billions of
dollars in ADAS and level one

through five autonomous
capabilities.

But those vehicle capabilities,
let's take an example of

the ADAS capability
of tailgating, right?

It says you can set up the
configuration to be one vehicle

distance, two vehicle,
three vehicle.

But as a consumer,

how do they know which
setting is safer for them?

Right?

There is a
recommendation, the OEMs,

but it's not personalized to
my safety characteristics.

So these AI foundational models
running right on the edge, right,

stores that safety characteristics
profile of the driver.

It can recommend what
setting is suitable for them.

Yeah. Right?

So I think the value of
coaching is not just about

giving immediate alert and

event notification,

but it's also about helping the
vehicle to be configured in a

way that is safe for them,

for their personalized
experience.

So we see that these AI
models not only helping the

insurance carriers assess
risk, the drivers improving

behaviors, how about
infrastructure entities like

DOTs and municipalities
wanting to understand what

type of behaviour is
happening in their counties,

sub counties, road network,
and how to improve it.

So all of them have to be able
to run on the edge, collect,

analyse the data.

These AI models run on the
edge, collect, analyse,

process, and those process data
goes and get distributed to the

different users based on the
UX experience they will like.

As you look forward
to the future,

where do you see embedded AI
and in vehicle intelligence

evolving next?

Just generally speaking,

obviously you're involved in
risk intelligence and so forth,

but maybe even beyond that.

My honest belief is that the
embedded AI is still in its

early stages, right?

And part of the problem
is the ability to manage

infrastructure
deployment, right?

We can build these
amazing models,

but we need a way to deploy
and get the value out to the

consumers, the customers, right?

I think right now

significant effort is
being put on safety,

let's say from
insurance perspective,

this risk assessment,
risk analysis models,

immediate driver
coaching, right?

I think the next step is

preventive

risk

routing capability.

For example, today,

I want to go from point A
to point B and I enter the

destination address, the system
then looks at congestion, you know,

estimated time of arrival
and a few other parameters to

recommend a route.

But imagine these AI
based analytical models

providing node level,
like, you know,

every intersection
point or every

mile, half a mile, whatever,

some certain fixed points, it
gives a scoring, right, risk score.

And if they get
sent to the cloud,

now your routing now can be
determined based on where

you're heading, right?

And then looking at, this route
has ten stop signs and many of

them had a high risk profile,

then let's route it in
another route, right?

So that's the next level of
routing innovation I believe

the edge AI will help.

And third one we already
touched upon is the

personalized experience
in the vehicle.

Today you can get in a vehicle
and maybe some of them have

like option one, two, three,

you select a preset memory
and that'll adjust your seat,

your steering wheel and maybe even
remember your music and so on, right?

This is fascinating.

Well, Siva, thank you for joining
us and sharing how MOTER is helping

reshape the future of Insurtech
and intelligent mobility.

This collaboration truly
highlights the growing

importance of edge
AI in the vehicle,

not only for enabling smarter
insurance and driver coaching,

but also for improving
privacy, scalability,

and real time responsiveness across
the connected vehicle ecosystem.

Thanks for listening and stay
tuned for more conversations

with innovators shaping the
future of software defined

vehicles and automotive AI.