Vention launches Montreal lab for industrial physical AI

Vention has opened a research laboratory in Montreal devoted to physical AI, the company said in an announcement reported by PRNewswire. The facility will investigate how robots can handle complicated, changing tasks in manufacturing, with work spanning industrial goods, electronics and automotive production.

Dr. Jimmy Li, Vention’s director of Physical AI, leads the lab. Dr. Joelle Pineau, Cohere’s chief AI officer, will advise the programme as an external technical adviser. Her role is expected to include input on model design, research priorities and links with the wider artificial-intelligence community.

The project is aimed at a practical problem: moving robotic capabilities from controlled demonstrations into factories where reliability, expense and changing conditions determine whether a system is useful. Vention says its researchers will test new methods against those production demands rather than treating laboratory performance as the final measure of success.

Research built around factory work

The lab’s programme brings together several areas of robotics and AI. These include the gathering of industrial data, robot control, planning for movement, traditional computer vision, vision foundation models, learning from demonstrations and reinforcement learning.

That combination is intended for manufacturing jobs that are difficult to structure in advance. A robot may need to recognise objects, understand their position, choose an appropriate grip and move without hitting nearby equipment. Such tasks become more demanding when products, layouts or working conditions vary.

Vention says the lab has already produced new intellectual property and will continue developing its work. Its researchers will have access to manufacturing environments through the company’s existing customer base, giving them a source of real-world data for further training of physical AI models.

The company says its technology is used by thousands of manufacturers worldwide, including 90 companies in the Fortune 500. Its wider platform has more than 28,000 machines deployed globally and serves a community of over 6,000 factories. Those figures give the lab a route into operating plants rather than relying solely on simulated or academic settings.

Vention describes the connection between research and customers as a continuing feedback cycle. Client responses, production difficulties and conversations about deployment can influence the technical work, while research results can be assessed against the requirements of actual manufacturing operations.

From models to deployable systems

Physical AI is currently Vention’s fastest-growing business area, with related revenue increasing by 400% during the past year, according to the company. Vention argues that expanding the use of these systems will depend not only on better models but also on reducing the complexity and cost of putting them into service.

The company’s stated goal is to make deployment accessible to a broader group of manufacturers as the underlying technology improves. That means building systems that can operate consistently, fit different production environments and provide sufficient economic value to justify their installation.

A major piece of the lab’s work is GRIIP, a modular physical-AI pipeline that Vention introduced in February 2026. The system covers several stages of robotic manipulation, beginning with creating a digital representation of a scene and continuing through object segmentation, pose estimation, grasp selection and collision-free movement planning.

GRIIP uses foundation models supplied by technology companies including NVIDIA alongside models developed by Vention. The company also plans to publish a software development kit for the system, allowing engineering teams to adapt and extend the pipeline for their own requirements.

Industry partnerships and leadership

Li joined Vention after research associated with McGill University and brings more than a decade of peer-reviewed work in computer vision, robot perception, autonomous systems and AI-based robotic manipulation. During his two years at Vention, he has led the company’s physical-AI strategy and its technical partnership with NVIDIA, with the aim of turning research advances into commercial systems for factories.

Vention is working with industrial and electronics manufacturers on demanding robotic applications. One project involves a large automotive original equipment manufacturer and focuses on complex, unstructured work in final assembly.

The company presents the new lab as part of its broader full-stack approach, combining modular hardware, software and physical AI. Its platform is designed to let businesses design, programme, install and run turnkey or customised automation systems. The Montreal centre is intended to strengthen the research layer behind that offering while keeping development connected to production needs.

For manufacturers, the significance of the effort lies in whether robotic systems can handle more varied work without requiring extensive custom engineering each time. Vention’s lab will attempt to close that gap by using its access to factories to shape, test and refine the technologies it develops.