Vention and its partners have invested in space and equipment for its new physical AI lab. Source: Vention
To generate high-quality data for training the AI models that operate next-generation robots, you need scale. Vention Inc. today opened its Physical AI Lab in Montreal. The company said the new laboratory focuses on robotic manipulation in manufacturing.
“Any physical AI foundation model is data-hungry,” Etienne Lacroix, founder and CEO of Vention, told The Robot Report. “We move several hundred robot cells a year, and they all collect high-quality industrial manipulation data. It was a massive asset that was not properly leveraged until now.”
Vention said its technology stack includes hardware, software, and physical AI. The Montreal-based company claimed that it enables businesses to design, program, and deploy turnkey or custom automation in days. It has deployed more than 28,000 machines worldwide and a community of more than 6,000 factories.
Vention addresses manufacturing challenges
With its technology deployed across manufacturers globally, including 90 of the Fortune 500, Vention said its scale gives researchers direct access to real environments and a continuous stream of data to post-train physical AI models.
“Physical AI will fundamentally expand what manufacturers can automate, but the challenge is no longer simply proving that a robot can perform a task in a lab,” stated Lacroix. “The real opportunity is making these capabilities reliable, economical, and deployable across thousands of factories. By combining Canada’s world-class AI ecosystem with Vention’s deep robotics expertise, industrial data, and full-stack automation platform, we have a unique foundation to close that gap.”
Revenue related to physical AI has increased by 400% over the past year, reported Vention. The company’s mandate is to enable scalable deployment by validating new AI capabilities against the reliability, cost, and variability requirements of real production lines.
“We know about Skild, Generalist, and Physical Intelligence — there are many models and policies,” Lacroix observed. “The base model is not the problem; some models are better for certain tasks. We can add value and shape the research agenda with efficient learning, data collection, and post-training capabilities.”
Editor’s note: Physical AI is among the session topic tracks at RoboBusiness 2026, which will be on Oct. 20 and 21 in Santa Clara, Calif. Register now to attend.
Register now and help us celebrate 20 years of RoboBusiness!Physical AI Lab takes research to the factory floor
Vention said its new lab includes industrial data collection, robotics control, motion planning, classical computer vision, vision foundation models, learning from demonstration, and reinforcement learning. It focuses on complex and unstructured manufacturing tasks.
The company noted that while model development is shaped by research and benchmarks, it also relies on industry feedback and day-to-day deployment conversations.
“What makes this lab different is the loop we’ve built: academic research feeding directly into live production problems, and production feedback feeding back into the research,” said Dr. Jimmy Li, director of physical AI at Vention. “Our clients aren’t waiting for a finished product to test; they’re in the room while we build it. That’s unique in this field, and it’s what lets us move faster from a research result to something that actually runs on a factory floor.”
Li, head of Vention’s Physical AI Lab, is a robotics and machine learning researcher from McGill University who has more than a decade of experience with peer-reviewed research in computer vision, robot perception, autonomous systems, and AI-powered robotic manipulation. Over the past two years, he has led Vention’s physical AI strategy and its collaboration with NVIDIA, translating advances in AI research into commercial robotic applications.
Jimmy Li, director of the new Physical AI Lab in Montreal. Source: Vention
Innovation under way for industrial applications
Vention said its new lab has already generated new intellectual property and plans to continue to do so. It launched GRIIP (Generalized Robotic Industrial Intelligence Pipeline) in February 2026. The modular physical AI pipeline spans scene digitalization, object segmentation, pose estimation, grasp selection, and collision-free motion planning.
GRIIP uses foundation models from technology leaders such as NVIDIA as well as Vention’s proprietary models. The company plans to release a public GRIIP Software Development Kit (SDK).
“We will put GRIIP in open source and show it alongside our other AI-defined automation at IMTS,” said Lacroix. Vention plans to introduce physical AI and agentic AI capabilities unified in a single platform.
The company is also working with global industrial and electronics manufacturers, including a large automotive OEM on high-complexity, unstructured robotic tasks in final assembly.
“We’ve always been in applied use cases, learning from demonstrations and data collection, using tens of millions of dollars in R&D, and fine-fine tuning models for industrial automation,” said Lacroix. “Kitting is one of our top applied research use cases, which is horizontal — it’s in automotive, in aerospace, and in consumer.”
“Electronics contract manufacturers have all those pieces of hardware coming from various vendors and various packaging, They need to be opened and reassembled into a kit that will go at a specific station on the assembly line,” he explained. “If you’re a car manufacturer, it could be headlamps with a harness, a mounting bracket, and a screw, and they need to be prepared so Position 7 on the line has the right kit for the right car. Kitting was a use case where those new class of policies are relevant.”
External advisor brings connections to physical AI community
In addition, Vention named Dr. Joelle Pineau, chief AI officer at Cohere, as external technical advisor for the Physical AI Lab. She will contribute to model architecture decisions, research prioritization, and connections into the broader AI community.
“Dr. Pineau was a former VP of research at Meta, and she did her Ph.D. at CMU,” said Lacroix. “She managed the teams that developed the SAM 2 model for segmentation.”
As foundation models mature, Vention said it expects the complexity and costs of deploying robots to drop, widening the range of manufacturers that can put them to work.
Tell Us What You Think!