AI NewsGo-to-marketReported
Qvyon raises $8M seed to test physical AI in dangerous edge cases
Qvyon came out of stealth with $8 million USD in seed money. The tool helps robot and self-driving truck makers test rare scenes that are hard to film.
Image: BetaKit
Why it mattersA self-driving truck can run safe for weeks and still fail on one scene nobody filmed. The cost of that gap is why teams buy edge-case data.
A self-driving truck can drive 10,000 kilometres and never see a child run into the road. It still has to work the one time it does. That gap, between what a camera recorded and what could happen, is what a Montréal startup wants to close. Qvyon Technologies came out of stealth with $8 million USD, or $11.4 million Canadian dollars, in seed money. BetaKit's Madison McLauchlan reports the round.
The round has $6 million USD of equity. The investors are Mac Venture Capital, BDC Capital's Black Entrepreneurs Fund, BKR Capital and Spotlight VP, per BetaKit. RBCx supplied $2 million in venture debt. Qvyon was founded by Frantz Saintellemy. He ran Canadian sensor company LeddarTech before this.
What the product does
Qvyon's product is called MIMIC Core. BetaKit describes it as a tool that blends real sensor data with an AI system. The AI then makes many variations from one recording. A team can film one scene on a real road. The tool spins out dark, wet and crowded versions of the same scene. The point is to give a team the hard scenes it never filmed.
Saintellemy told BetaKit why the problem exists. He said the hard scenes were difficult to anticipate and expensive to capture. He also said that some were sometimes dangerous to reproduce, and sometimes impossible to recreate in nature.
Who is using it
BetaKit names two Qvyon customers. The first is NuPort Robotics. It is an autonomous trucking company in Markham, Ontario. The second is Voltava. It is a US maker. Its Qvyon work is on factory inspection systems. Qvyon has about 20 people on staff. The company runs out of Montréal.
Saintellemy told the outlet he sees room to scale fast. He said he wants to build a foundational company in Canada. The company's roadmap includes two more products for 2027, per BetaKit. One measures test results. The other improves the AI on data a robot sees in the field.
Why it reads as a real story
The seed stage is early. A company with 20 people and two named customers is still a small bet. The sign of a real bet is the money. Three venture funds and a bank's venture arm put in the round. The founder has run a larger sensor business before. The problem is one the industry writes about openly. Teams do not have the long-tail data they need. Real-world capture is slow and expensive.
For a founder or a buyer of AI software, the story points at a working question. If your product acts in the real world, you need test data for scenes you have never seen. Companies like Qvyon now sell that data as a product. Whether the synthetic scenes catch real failures is a test the next 18 months will run.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
