Research labs producing AI training data

Name: Research labs producing AI training data
Creator: Blomega
License: https://creativecommons.org/licenses/by/4.0/

39 records, built 2026-09-18. Machine-readable: JSON and JSON-LD.

Name	Institution	Country	Focus	Grade	Source
1X Technologies	1X Technologies	Norway/USA	consumer humanoid, world model	A	source
A2I (Oxford Robotics Institute)	University of Oxford	UK	robot learning, learning from demonstration	B	source
AgiBot (Zhiyuan)	AgiBot	China	mass-produced humanoids, largest open dataset	B	source
Agility Robotics	Agility Robotics	USA	bipedal logistics robot, teleop-then-autonomous	B	source
Apptronik	Apptronik	USA	industrial humanoid, Robot Park data factory	A	source
AUTOLAB	UC Berkeley	USA	grasping, cloud robotics, manipulation	B	source
Cognitive Robotics (Kober group)	TU Delft	Netherlands	interactive robot learning, imitation from few demos	B	source
Dyna Robotics	Dyna Robotics	USA	world-action model, scaling via human video	A	source
Encord	Encord	UK/USA	data layer for Physical AI: multimodal/point-cloud (data vendor)	A	source
Figure AI	Figure AI	USA	humanoid VLA (Helix), consumer, industrial	A	source
Fourier Intelligence	Fourier Intelligence	China	humanoid, care robots, open-source N1	B	source
Generalist AI	Generalist	USA	embodied foundation models, general-purpose robots	A	source
Google DeepMind Robotics	Google DeepMind	USA/UK	vision-language-action models at scale	A	source
GRAIL	New York University	USA	large-scale robot learning, affordable open robots	B	source
Improbable AI Lab	MIT CSAIL	USA	sim-to-real, dexterous manipulation, RL	A	source
IRIS Lab	Stanford University	USA	robot learning, imitation & RL, manipulation foundation models	A	source
IRoM-Lab	Princeton University	USA	safe generalization, formal guarantees, manipulation	A	source
MVIG	Shanghai Jiao Tong University	China	general robotic grasping/manipulation, vision, haptics	A	source
NVIDIA GEAR (Isaac GR00T)	NVIDIA	USA	open humanoid foundation model, sim	A	source
Perceptual Science Group (GelSight)	Massachusetts Institute of Technology	USA	vision-based tactile sensing for manipulation	A	source
Physical Intelligence	Physical Intelligence Inc.	USA	cross-embodiment VLA foundation models	A	source
R-PAD	Carnegie Mellon University	USA	self-supervised learning, generalization, manipulation	C	source
RAIL (BAIR)	UC Berkeley	USA	deep RL, offline RL, VLA models for control	A	source
REAL	Stanford University	USA	manipulation via interaction, diffusion policies	A	source
RL2	Georgia Tech	USA	visuomotor skill learning, imitation, world models	A	source
Robot Learning Lab	Imperial College London	UK	dexterous manipulation, sim-to-real, one-shot imitation	A	source
Robotic Systems Lab (RSL)	ETH Zurich	Switzerland	legged locomotion, loco-manipulation	A	source
RPL	University of Texas at Austin	USA	generalist robots, imitation/RL, sim frameworks	A	source
RSE-Lab	University of Washington	USA	state estimation, perception, manipulation	A	source
Sanctuary AI	Sanctuary AI	Canada	teleop-collected behavioral, tactile data	A	source
Scale AI (Physical AI)	Scale AI	USA	teleop capture, sensor annotation, eval (data vendor)	A	source
Skild AI	Skild AI	USA	omni-bodied robot foundation model	A	source
SVL	Stanford University	USA	vision, manipulation, embodied AI benchmarks	C	source
Tactile Robotics Group (BRL)	University of Bristol	UK	vision-based tactile sensing, tactile sim-to-real	A	source
TEA Lab	Tsinghua University	China	embodied AI, visual RL, manipulation	B	source
Tesla Optimus	Tesla	USA	factory humanoid, vision-only data collection	B	source
Toyota Research Institute	Toyota Research Institute	USA	Large Behavior Models, diffusion policy, dexterity	A	source
UBTech Robotics	UBTech	China	industrial humanoids, embodied-AI data centers	B	source
Unitree Robotics	Unitree	China	low-cost humanoids/quadrupeds, open teleop tooling	A	source

Cite as: Blomega Data Refinery, https://data.blomega.com, CC BY 4.0. Published by Blomega (Wikidata Q141048865). Unknown values are shown as not established, never guessed.