TP7's Mobile Dual-Arm Robot Rolls Off the Pallet Jack and Into the High-Mix Factory

The Vancouver startup, backed by accelerators from NVIDIA to Harvard, is betting its physical-spatial AI can automate messy, variable tasks without redesigning the floor.

About TP7

Published

The most expensive part of a factory robot is often the factory. Conveyor belts get ripped out, cages are welded into place, and entire production lines are redesigned to accommodate a single, inflexible arm. TP7, a Vancouver-based startup, is making a different bet: that a mobile, dual-arm robot powered by what it calls physical-spatial AI can simply roll off a pallet jack and start working, learning new tasks by watching and listening in the messy, high-variance environments where legacy systems fail [tp7.ai, 2025].

High-mix manufacturing and logistics have largely resisted automation. The unit economics of a stationary robot programmed for one job collapse when that job disappears tomorrow. TP7's answer is a robot built for variability, claiming 24/7 operation with no safety cages required [tp7.ai, 2025]. The company is affiliated with a roster of accelerators, including NVIDIA Inception, Mass Robotics, Harvard Innovation Labs, and SFU VentureLabs [venturelabs.ca, 2026] [massrobotics.org, 2026] [innovationlabs.harvard.edu, 2026].

The Wedge: Drop-In Automation

The core promise is 'drop-in automation, zero redesign' [tp7.ai, 2025]. Instead of a multi-month engineering project, the idea is to unload a TP7 robot, give it a natural language instruction or a demonstration, and have it start picking, packing, or kitting within hours. The technology stack leans on AI for real-time reasoning about physical space, sensor fusion, and learning from demonstration [tp7.ai, 2025].

Public profiles also mention dual-use applications in defense and emergency response, where robots might navigate unstructured disaster zones or handle hazardous materials [venturelabs.ca, 2026].

An Early-Stage Bet on Proprietary AI

The company was founded in 2024 and lists a Boston presence in some accelerator profiles, though its headquarters are in Vancouver [pitch.vc, Oct 2024]. A CTO, Hadley Fox, is listed in association with the company, but full team details, customer names, and funding amounts are not disclosed [LinkedIn, 2026]. The company exhibited its 'drop-in automation' concept at the Hannover Messe industrial fair [The Logic, 2025].

  • Technical frontier. 'Physical-spatial AI' is not a standard term. Delivering human-like dexterity and reasoning in constantly changing environments is an unsolved problem at scale.
  • Commercial path. The dual-use focus could stretch early resources. Defense contracts offer large potential value but involve long, complex sales cycles.
  • The data moat. The company says its robots learn via natural language and demonstration [innovationlabs.harvard.edu, 2026]. The value of that claim hinges on a proprietary dataset of physical interactions that competitors cannot easily replicate.

If a traditional automation line requires two weeks of installation and calibration, and a TP7 robot truly needs just four hours, the payoff is the flexibility to reallocate automation on a daily basis. To win, TP7's machines must be not only capable, but also profoundly easy to live with.

Read on Startuply.vc