Inverted AI raises $5.3M to advance self-driving cars

The company is at the forefront of solving one of autonomous vehicles' biggest challenges.

Photo: Shutterstock

Creating simulations that match reality is one of the most complex roadblocks to making autonomous vehicles (AVs) safer. Even Elon Musk highlighted that this undertaking would be a monumental achievement. But progress is being made close to home through Inverted AI, a research and product company spun out of UBC’s Pacific Laboratory for Artificial Intelligence (PLAI), and a graduate of the university’s Creative Destruction Lab (CDL). Inverted just closed its CAD $5.3 million seed funding led by Vancouver’s Yaletown Partners to build on and expand its product offerings — which significantly cuts the time and costs required for AV development.

Why simulations are critical to AVs

Simulations are used to test, validate, and train autonomous vehicles, and could reduce the testing miles required to show proof-of-safety. The industry’s challenge is predicting the behaviour of “agents” — anything moving around vehicles like other cars, pedestrians, and cyclists  — to prevent overlaps that lead to collisions. 

“The main thing that people would do is manually program all these scenarios,” said Frank Wood, CEO and co-founder of Inverted. 

It’s impossible, of course, to do this for every human behaviour, yet it’s critical information for the technology to make safe decisions. For simulations to be effective for AVs, agents need to have diverse, reactive, and realistic behaviours — essentially, to be human-like. 

To fill this need, Inverted uses deep learning and generative AI to understand and anticipate driver, cyclist, and pedestrian behaviours. Inverted’s predictive model learns from massive quantities of video data from drones all over the world. The models are used to offer two available products for AV development: Initialize and Drive. Initialize produces the agent placements in a simulation, and Drive enables agents to have diverse, reactive, and realistic behaviours.

Inverted is now beta-testing two new products: Blame and Scenario. Blame automatically determines which agents caused a collision and why, and Scenario enables a whole-scene generator with agents that include buses and lights.  

“We provide value to AV companies by producing better simulators than they have, and that's what we want,” highlighted Wood. ”But our quest is to solve this overlapping problem [of agents] perfectly […] We do that better than anyone else.” 

Inverted’s top-notch team 

Wood added that beyond Inverted having a genius co-founder, Adam Ścibior, the company has naturally attracted a diverse team — one that is both gender-balanced and includes representatives from a number of countries — and that it’s one of the things that makes him most proud.

I think having built an organization of people that really care, [who] are drawn from all over the place, and being able to achieve that in Canada slash Vancouver — it feels good and effective in the right way,” said Wood.

Along with recognizing Inverted’s impressive team, Yaletown Partners became a strategic investor for a few reasons. The Vancouver-based venture capital firm is keen to support opportunities for academic research to be commercialized, as well as graduates of UBC’s CDL more generally.

Besides Inverted, Yaletown has also invested in InnerSpace, which designs and analyzes how to arrange office spaces;, a platform that lets developers run and test code without having to manage servers or infrastructure; and Elevated Signals, a process-management platform for agriculture and manufacturing.

“I think Inverted is coming out as the ground-floor foundation to support the growing Vancouver AI ecosystem,” said Eric Bukovinsky, a partner at Yaletown.

Following the semiconductor path

Bukovinsky added that Yaletown is also focused on how AI is being applied within industry processes and functions. The technology, he says, should be viewed as a resource that has the potential to follow the same path as semiconductors did with creating compute resources.

“Now, the cost of us to write a very complex program can be put on your credit card, rather than having to buy a gigantic system in a warehouse somewhere,” said Bukovinsky. “You're at the front end of this major process and it's key to support these types of things.”

Colin Harris, an Inverted board member, agrees with Bukovinsky’s perspective. Harris is the co-founder of PMC-Sierra, now renamed Microchip, a Nasdaq-listed semiconductor company that contributed to the evolution of the internet.

Harris explained that about 20 years ago, more than half the cost of chip development went towards making sure they worked. “I think [AVs] are in a similar boat,” said Harris. “Maybe the original people doing it didn't understand the cost that they would have to pay to do that. We had this problem in the chip world [...] and then we solved it,” adding that Inverted’s team is capable of doing the same with AVs.

For Harris, Inverted's approach can be used in a lot of areas, and is a big step forward on the AI journey. “This technique is probably the next big thing, for sure.”

Don't miss out on the funding and deals happening in your backyard. Subscribe or become a member of Vancouver Tech Journal

Join the conversation

or to participate.