Your Vehicle is about to be a lot Smarter

Announcing our investment in Viaduct

By: Harpinder Singh and Davis Treybig

Matej Drev (left), Head of Business Development and David Hallac (right), Founder and CEO

Have you ever wondered why your car dealership still tells you to get your car serviced every 5000 miles, independent of how you drive or what is actually going on in your car? What about the fact that to get usage-based vehicle insurance, you typically need to install a third-party measurement device to collect the data? In a world where intelligent, data-driven products increasingly permeate everything we do, it seems that the vehicles we drive, both personal and commercial, have been left behind. These basic use cases seem totally out of reach, and the really exciting stuff, such as a vehicle that automatically personalizes its adaptive drivings systems based on who you are and how you’re driving today, seems like a pipe dream.

The cause of this is simple: vehicles have historically not been “connected”. In other words, there has been no continuous data transfer between the vehicle and the cloud. While some vehicles may have sent a few bits of information here and there, it was not close to comprehensive enough to allow for exciting features and functionality to be built on top.

The good news is that this has notably begun to change in the last 4 or so years — automotive OEMs and telecommunications providers have finally built the data infrastructure needed to connect hundreds of millions of vehicles to the cloud. The bad news is that this has only led to a second, arguably greater challenge — figuring out how to make sense of and leverage this enormous amount of data to improve the lives of the billions of people who rely on these vehicles every day.

As an example, consider the challenge of predicting automotive part failures. Anticipating failures before they occur saves time, money, and stress for drivers on the road, makes vehicles safer, and has a profound impact on utilization in commercial vehicles like trucks. However, a typical vehicle today has tens of thousands of “DTC’s” or diagnostic codes that can occur at any given time, alongside regularly collected telematics data about things like speed, turn angle, and depression of the brake pad. A typical OEM might be collecting a snapshot of these signals every few seconds across a fleet of tens of millions of vehicles. Imagine now trying to use that data to predict a part failure that occurs on the order of one out of every ten thousand vehicles each month. You’re essentially trying to find a needle in a haystack, and it only gets worse if you’re an OEM that has traditionally focused on hiring manufacturing talent, not data scientists.

This is a hard problem and one that our friend David Hallac spent years researching during his Ph.D. thesis at Stanford. David’s work was focused on massive scale, multivariate time-series analysis — figuring out how to merge countless different types of time-series signals and model them to produce accurate, high fidelity predictions. His PhD research was so profound that many Fortune 100 companies actively sought him out at Stanford to apply his techniques to their datasets, ultimately revealing the impact his research might have on the automotive industry at large. As such, after graduating, David decided to fulfill that vision and start Viaduct in order to provide the intelligence layer so desperately needed by the automotive industry. Jure Leskovec, David’s PhD advisor at Stanford and a world-class expert in machine learning, joined him as a founding advisor to the company. Since then, they’ve assembled a superb team that blends both automotive expertise and deep technical aptitude perfectly suited to go after this problem.

It’s rare to find someone so technically apt who also has such a strong grasp on a market, but David effortlessly blends both. We’ve had the fortune of knowing David since his earliest days starting Viaduct, and have been blown away by his success breaking into many of the largest commercial and passenger automotive OEMs in the world as such a small startup. Viaduct’s analytics engine is already providing insights for hundreds of thousands of trucks and passenger vehicles worldwide and demonstrating extraordinarily high precision predictions. This early success lays the foundation for a broader suite of products across predictive maintenance, personalized in-vehicle experiences, usage-based insurance, and so much more. As vehicle ownership increasingly moves away from self-ownership and towards managed, and eventually self-driving, fleets, the importance of this technology will only be more profound.

As such, we couldn’t be more excited to lead Viaduct’s Series A round. This investment extends our firm’s longtime focus on businesses leveraging breakthroughs in data science to transform large industries, such as the life sciences, food and agriculture, supply chain, the industrial domain, and more. It also highlights our continued efforts to fund companies built on top of profound academic research. We can’t wait to see how the vehicles of the world transform as a result.

Investing in visionary founders, transformational technology and emergent ecosystems for a new world.

Recommended from Medium

Data Science for Medical Imaging and the Business Behind it

Google Fuzzing: Is the privacy/utility tradeoff a false dichotomy?

[ML UTD 8] Machine Learning Up-To-Date — Life With Data

Hayden AI Welcomes Seasoned Product and Supply Chain Leader Jeff Nielsen as VP of School…

Clear Conquered U.S. Airports. Now It Wants to Own Your Entire Digital Identity.

The Art of Neural Networks

O2C: Achieving Process Excellence in SSC through E2E Solutions and Technologies

RetailTech. Meet HardTech.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Harpinder Singh

Harpinder Singh

Innovation Endeavors

More from Medium

The Downsides of Algorithmic Choices

Maturity Levels of OR Projects

Bias Variance Tradeoff

png

Feature Selection using FRUFS with Classification using K-Means and VevestaX