FactoryIntelligence builds industrial robots that combine a multimodal tactile foundation model with a hard‑real‑time neuroclassical control layer, enabling anticipatory contact reasoning and safe force compliance. Their purpose‑built sensors, grippers, and unified software runtime collect and share operational data across fleets, so improvements from a single robot are instantly applied to all units.
Funding
Funding not disclosed
Founders
Product
Problem
Industrial robots often rely on vision‑only perception and generic control loops, which limits their ability to handle delicate, contact‑rich tasks safely and reliably. Without integrated touch sensing and real‑time compliance, robots can damage parts, fail to adapt to variations, and require extensive hand‑tuning for each new application.
Solution
FactoryIntelligence addresses this gap by combining a tactile foundation model with a classical control layer, purpose‑built hardware, and unified software orchestration. The foundation model is trained on vision, touch, proprioception, and language, enabling the robot to anticipate and reason about contact events. A low‑level neuroclassical controller enforces force and compliance limits in hard real time, providing safety even when the learned model errs. Custom sensors, grippers, and actuators are designed to deliver accurate force feedback that the model can trust. A single runtime sequences skills across fleets, streams operational data back to the training loop, and propagates learning from one robot to all others, accelerating deployment of contact‑aware automation.
Target Audience
Target customers are manufacturers and automation integrators that need robots capable of delicate assembly, handling, or inspection tasks where tactile feedback and real‑time compliance are critical.
Features
- Multimodal foundation model integrating vision, tactile, proprioceptive, and language inputs for anticipatory contact reasoning
- Neuroclassical control layer that enforces force and compliance limits in hard real time, ensuring safety independent of the learned model
- Purpose‑built sensor suite and compliant grippers calibrated to the model’s data requirements
- Unified runtime that orchestrates skill execution, coordinates multiple robots, and streams telemetry for continuous learning
- Fleet‑wide knowledge sharing where data from a single robot updates the model for all deployed units