The new facility aims to bridge the gap between theoretical machine learning research and the rigorous demands of large-scale industrial manufacturing environments.
Vention officially inaugurated its Physical AI laboratory in Montreal on September 9, 2026, establishing a dedicated research hub for integrating machine learning into industrial robotic manipulation. The facility prioritizes the transition of advanced AI research into production-grade systems capable of operating within the complex, unstructured environments characteristic of modern manufacturing.
Dr. Jimmy Li, the director of the new laboratory and a former researcher at McGill University, leads a multidisciplinary team currently composed of eight specialists. The research agenda focuses on the intersection of robotics control, motion planning, and vision-based foundation models. This technical program incorporates reinforcement learning and learning from demonstration to address high-complexity tasks in sectors such as automotive assembly and electronics production.
The laboratory functions as a feedback loop between theoretical development and industrial application. By leveraging Vention’s existing infrastructure, which includes thousands of deployed machines across 90 Fortune 500 companies, researchers gain access to high-fidelity industrial data. This data stream serves as the foundation for post-training models, ensuring that algorithmic improvements are validated against the specific reliability and cost constraints of factory floors.
A central component of the lab’s output is the GRIIP pipeline, a modular architecture introduced in February 2026 that manages scene digitalization, object segmentation, and collision-free motion planning. The system utilizes a combination of proprietary Vention models and foundation models from technology partners like NVIDIA. Plans for a public Software Development Kit are underway, which will allow external engineering teams to iterate on the pipeline for specialized use cases.
The lab is currently expanding its headcount to support these initiatives, with over 16 open positions targeting expertise in simulation, robotics control, and Physical AI research. This recruitment drive reflects a strategic commitment to scaling the development of autonomous systems that can adapt to the variability inherent in high-speed production lines. By focusing on both hardware integration and software optimization, the team aims to reduce the latency between model training and real-world deployment.
Dr. Joelle Pineau, Chief AI Officer at Cohere, serves as the External Technical Advisor for the initiative. Her role involves providing strategic guidance on model architecture and research prioritization as the lab scales its operations. Her involvement signals a focus on aligning industrial robotics with the latest advancements in large-scale foundation models and multimodal AI architectures.
The economic rationale for the lab centers on reducing the complexity costs associated with deploying AI in industrial settings. Etienne Lacroix, founder and CEO of Vention, noted that the primary challenge in the field remains the transition from controlled laboratory demonstrations to consistent performance in real-world production environments. By embedding the research process within a live industrial ecosystem, the company seeks to standardize the deployment of intelligent robotics at scale.
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. 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.
The integration of academic rigor with industrial feedback mechanisms distinguishes this facility from traditional robotics research centers. Dr. Jimmy Li emphasized that the proximity to production challenges allows for rapid iteration cycles that are often absent in purely theoretical environments. This methodology ensures that the research output remains tethered to the practical requirements of end-users who require robust, repeatable automation.
Future milestones for the laboratory involve the continued refinement of vision foundation models for robotic perception and the expansion of the GRIIP SDK. Stakeholders will monitor the lab’s ability to maintain performance benchmarks as the complexity of the automated tasks increases. The success of this initiative will likely influence how manufacturers approach the integration of generative and predictive AI into their existing automation stacks.
