In an era where technology is rapidly evolving, Vention Inc. has taken a significant step forward in revolutionizing the manufacturing sector. The company has recently announced the commercial launch of its Rapid Operator AI at NVIDIA GTC 2026, a system designed to autonomously perform complex, unstructured tasks that typically involve deep picking. The AI-based system identifies and grasps randomly oriented parts in dense containers using state-of-the-art perception and motion planning.
Revolutionizing Manufacturing with Rapid Operator AI
Unlike other AI solutions designed for structured environments like warehouses, Rapid Operator AI is specifically built to tackle unstructured manufacturing tasks. As Vention’s founder and CEO, Etienne Lacroix, points out, “The world of manufacturing is much more demanding.” The new product is built on the company’s Generalized Robotic Industrial Intelligence Pipeline (GRIIP), which integrates Vention’s proprietary models with open NVIDIA Isaac models.
Targeted at mid-sized and enterprise manufacturers operating multi-shift facilities, Rapid Operator AI aims to alleviate operational tension created by labor shortages and high production variability. The technology is poised to revolutionize the way these manufacturers operate, providing a unique solution to a common problem.
Why Deep Picking?
Vention’s decision to focus on deep picking tasks was driven by two key factors. First, their customers indicated this was a recurring issue in their operations. “In machine assembly or maintenance, you have a bin of parts, and then you have to take them out of the bin and then do an operation with them,” explains François Giguère, Technology Director at Vention.
Secondly, the complexity of deep picking tasks presents a unique challenge that, if overcome, could unlock solutions for other problems in manufacturing. This task involves the careful picking of parts from deep bins, a process that requires accurate perception and collision-free motion planning to ensure the robot doesn’t collide with the bin or other objects inside it.
Building an Efficient and Flexible AI Model
Rapid Operator AI boasts an impressive range of capabilities. It can detect randomly oriented parts in dense clutter, estimate accurate poses at 6-DoF (degrees of freedom), and plan collision-free picks. The system can also handle containers up to 24 inches deep, and can operate in various lighting conditions and materials, offering a level of adaptability that is crucial in diverse manufacturing environments.
To achieve this, Vention leveraged the best parts of the world’s AI pipelines and models. “You want generalizability, and you want speed and performance,” Lacroix said, highlighting the need for AI models that are not only efficient but also flexible enough to meet diverse manufacturing needs.
The Role of NVIDIA in Development
Contributing to the success of Rapid Operator AI is NVIDIA’s FoundationStereo and FoundationPose technologies, which Vention utilized for stereo matching and pose estimation. The integration of these technologies into Vention’s system enabled the creation of an efficient and powerful AI model capable of tackling complex manufacturing tasks.
Moving forward, Vention intends to keep Rapid Operator AI focused on manufacturing tasks, while the GRIIP foundation allows for a wider variety of tasks to be covered. Lacroix concludes, “Any manufacturer that operates a two-shift factory can now deploy physical AI with a two-year payback. You get the speed of humans, the reliability of humans in terms of selection, and you’re able to navigate, at the same time, these very complex, very constrained manufacturing environments without any collisions.”
Source: Here
