Ex-Meta Scientists Launch Industrial Vision AI Model Isaac 0.5 With $21M in Funding

Perceptron, a startup founded by two former Meta researchers, launched its latest AI model in August 2026 aimed at helping robots navigate warehouses and factory floors. The model, called Isaac 0.5, is designed to give machines the ability to perceive, reason, and act in industrial settings — and is being released as an open-weight model, meaning anyone can inspect its parameters and training materials.

The company was co-founded by Armen Aghajanyan and Akshat Shrivastava, who previously worked at Meta’s Fundamental AI Research (FAIR) division. Perceptron was founded in November 2024 and recently raised $21 million in a funding round led by Bessemer Venture Partners.

Unlike narrower industrial AI tools built for single, repetitive tasks, Perceptron says Isaac 0.5 is general-purpose and flexible across different environments and situations. Shrivastava used package sorting as an example of the complexity involved: a robot must read labels, analyze spatial layouts, decide which box to pick up, and plan the sequence — steps that current software handles separately rather than as a unified, adaptable process.

Isaac 0.5 was trained on a million hours of general video, along with ego video — footage captured from a first-person perspective of a person completing a physical task — and UMI video, which records repetitive human actions to teach AI movement. Shrivastava said the company has built “internally petabyte-scale datasets that span across modalities, whether it’s images, text, video, etc. all the way through robotic trajectories,” though specific data sources were not disclosed.

Perceptron says it is targeting customers across manufacturing, logistics and warehousing, security, mobility, and media and entertainment. The open-weight release of Isaac 0.5 could allow the model’s intelligence layer to be integrated across a broad range of industries and vendors.