Field Foundation Models (Field AI)

Field AI's robot foundation model, built to operate autonomously in unmapped construction sites and industrial facilities.

FFM (Field Foundation Models) is the core model line of Field AI, a US robot-software company headquartered in Irvine, California. Its founder and CEO is Ali Agha, and the team includes people from NASA's Jet Propulsion Laboratory and Google DeepMind, with experience on Mars exploration and DARPA programs. The company describes FFM as a “physics-first” embodied foundation model: it combines data-driven neural networks with physics-based reasoning and uncertainty estimation, built around a risk-assessing “belief world model” that lets a robot move and work autonomously in unstructured environments with no pre-built map, no GPS, and no preset path — running entirely on onboard compute. The same model can be installed on robots of different shapes from different manufacturers, which the company describes as “one brain for any machine.” According to its press releases, Field AI has raised more than $400 million in total funding, from investors including Bezos Expeditions, Khosla Ventures, and NVIDIA's venture arm NVentures.

ExampleIn March 2026, Field AI announced a partnership with Boston Dynamics to put FFM on the Spot quadruped, letting it move and work autonomously on construction sites that have no map and change day to day; that September, the company also announced a partnership with Caterpillar.

Also called

FFM

Related

Field AI · Foundation Model · World Model · Uncertainty Estimation · Boston Dynamics Spot · On-Device / Edge Deployment

Sources

FieldAI 官网 (Chinese)
Boston Dynamics and FieldAI Partner to Bring Robots Into Uncharted and Dynamic Environments
Caterpillar and FieldAI Advance AI-Powered Industrial Innovation (PR Newswire, 2026-09-02)

As of

2026-09