# Robotics in the San Francisco Bay Area

> Source: https://aiwiki.ai/wiki/robotics_in_the_san_francisco_bay_area
> Updated: 2026-10-13
> Categories: AI Research, Robotics
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Robotics in the San Francisco Bay Area." aiwiki.ai, 13 Oct 2026. https://aiwiki.ai/wiki/robotics_in_the_san_francisco_bay_area
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

Robotics in the San Francisco Bay Area includes academic research in perception and control, companies developing robot hardware and learned policies, and suppliers of computing and simulation tools.[1][2][18] [Stanford University](https://aiwiki.ai/wiki/stanford_university) and the [University of California, Berkeley](https://aiwiki.ai/wiki/uc_berkeley) conduct [AI robotics](https://aiwiki.ai/wiki/ai_robotics) research, while local companies work on humanoids and general-purpose robot models.[1][2][3][4][6][7] These activities span several locations rather than a single campus or industrial park.[3][6][8]

## Geographic scope

The Metropolitan Transportation Commission uses a nine-county definition of the San Francisco Bay Area.[5] Within that region, robotics organizations have locations in Berkeley, San Francisco, Stanford, Palo Alto, Santa Clara, San Jose and Menlo Park.[1][2][3][6][8][9][10][11] The locations below identify the presence stated by each organization, not a claim that its whole operation is based in the region.

## Early AI robotics

SRI's Artificial Intelligence Center developed Shakey between 1966 and 1972. The mobile robot combined perception with reasoning about its surroundings and performed tasks involving route planning and the movement of simple objects. Shakey provides an early example of the region's research connecting AI planning to a physical machine. SRI lists its address in Menlo Park, and Shakey is now housed at the Computer History Museum.[10]

## University laboratories

Stanford opened its Robotics Center in November 2024 in the Packard Electrical Engineering Building.[1][20] Its facilities include a domestic suite with a kitchen and laundry equipment, and a medical suite with a surgical robot. The center brings together research on manipulation, human interaction and other robotics applications. Its laboratory directory includes the Interactive Perception and Robot Learning Lab, Intelligence through Robotic Interaction at Scale Lab, and Robotics and Embodied Artificial Intelligence Lab.[1][11]

At Berkeley, the Berkeley Artificial Intelligence Research (BAIR) Lab includes researchers in computer vision, machine learning, planning, control and robotics. The Robotic AI & Learning Lab studies algorithms for machines to acquire flexible behavior through learning. Berkeley's InterACT laboratory focuses on human-robot and human-AI interaction, including learning from human feedback and modeling human decisions. These programs address both robot capabilities and how robots should act around people.[2][12][13]

## Robot learning research

[Robot learning](https://aiwiki.ai/wiki/robot_learning) is one connection between these universities and industry. Published projects cover demonstrations collected by human operators, policies that convert observations into motor actions, and adaptation to unfamiliar environments. Their experiments use specified robot setups and tasks, rather than an unrestricted test of general intelligence.[14][15][16]

### Learning from demonstrations

The 2023 [ALOHA](https://aiwiki.ai/wiki/aloha_robot) project, by researchers affiliated with Stanford, Berkeley and Meta, paired a bimanual teleoperation system with [Action Chunking with Transformers](https://aiwiki.ai/wiki/action_chunking_transformer) (ACT). ACT learns from demonstrations and predicts sequences of robot actions instead of one action at a time. The study tested fine manipulation such as opening a condiment cup and inserting a battery. It is an example of [imitation learning](https://aiwiki.ai/wiki/imitation_learning) applied to physical skills using relatively inexpensive robot hardware.[14]

### Vision-language-action models

[OpenVLA](https://aiwiki.ai/wiki/openvla), introduced in 2024, is a 7-billion-parameter [vision-language-action model](https://aiwiki.ai/wiki/vision_language_action_model) trained on 970,000 real-world robot demonstrations. Its authors include researchers affiliated with Stanford, Berkeley, Toyota Research Institute, Google DeepMind, Physical Intelligence and MIT. The project released model checkpoints and training code. Its evaluations examine manipulation across multiple robot setups and fine-tuning for new tasks, including the use of parameter-efficient adaptation.[15]

### Distributed data collection

The 2024 DROID project collected 76,000 demonstration trajectories across 564 scenes through a collaboration spanning North America, Asia and Europe. It released the dataset, policy-learning code and instructions for reproducing its hardware. The portable setup uses a Franka Panda arm and teleoperation controls. DROID illustrates that robotics data used by researchers connected to the region can come from a much wider geographic network.[16] OpenVLA's project also evaluates adaptation to the DROID robot setup.[15]

## Companies and infrastructure

The following examples distinguish local presence from an organization's technical role. They are not a census of the region's companies.

| Organization | Verified regional presence | Robotics work |
| --- | --- | --- |
| [Figure AI](https://aiwiki.ai/wiki/figure_ai) | Headquarters in San Jose | Develops general-purpose humanoid robots, including the Figure robot series.[3][4] |
| [Physical Intelligence](https://aiwiki.ai/wiki/physical_intelligence) | Office in San Francisco's Mission District | Develops robot foundation models and learning algorithms; publishes the openpi model and training repository.[6][17] |
| [Generalist AI](https://aiwiki.ai/wiki/generalist_ai) | Bay Area location, alongside a Boston location | Develops embodied foundation models with an initial focus on dexterity.[7] |
| [NVIDIA](https://aiwiki.ai/wiki/nvidia) | Corporate location in Santa Clara | Supplies computing, simulation, robot-learning software and AI models for robotics.[8][18] |
| [Tesla Optimus](https://aiwiki.ai/wiki/tesla_optimus) program | Optimus inference-hardware role listed in Palo Alto | Develops a general-purpose bipedal humanoid robot; its hardware work includes onboard AI inference.[9][19] |

Robot platforms and robot models are separate products. Physical Intelligence's openpi repository, for example, provides pretrained models and examples for adapting them to datasets and robot platforms. The project warns that a model trained for one setup may not work successfully on another. Such adaptation requirements limit what can be inferred from a demonstration of a single robot.[17]

NVIDIA's tools address another part of development. [Isaac Sim](https://aiwiki.ai/wiki/nvidia_isaac_sim) provides physically based simulation and virtual robots and sensors. [Isaac Lab](https://aiwiki.ai/wiki/isaac_lab) adds environments and workflows for training and evaluating policies using reinforcement learning and demonstrations. These tools support development by organizations beyond the Bay Area as well as those with local offices.[18]

## References

1. Stanford Engineering. ["New center unites Stanford's robotics expertise under one roof"](https://engineering.stanford.edu/news/new-center-unites-stanfords-robotics-expertise-under-one-roof). Accessed October 13, 2026.
2. Berkeley Artificial Intelligence Research. ["About"](https://bair.berkeley.edu/about.html). Accessed October 13, 2026.
3. Figure. ["Careers"](https://www.figure.ai/careers). Accessed October 13, 2026.
4. Figure. ["Company"](https://www.figure.ai/company). Accessed October 13, 2026.
5. Metropolitan Transportation Commission. ["What Is MTC?"](https://mtc.ca.gov/about-mtc/what-mtc). Accessed October 13, 2026.
6. Physical Intelligence. ["Join us"](https://www.pi.website/join-us). Accessed October 13, 2026.
7. Generalist. ["About"](https://generalistai.com/about). Accessed October 13, 2026.
8. NVIDIA. ["Contact NVIDIA"](https://www.nvidia.com/en-us/contact/). Accessed October 13, 2026.
9. Tesla. ["Electrical Engineer, SI Inference Hardware, Optimus"](https://www.tesla.com/careers/search/job/electrical-engineer-si-inference-hardware-optimus-264544). Accessed October 13, 2026.
10. SRI. ["Shakey the Robot"](https://www.sri.com/hoi/shakey-the-robot/). Historical project, 1966-1972.
11. Stanford Robotics Center. ["Labs"](https://src.stanford.edu/labs). Accessed October 13, 2026.
12. UC Berkeley Robotic AI & Learning Lab. [Laboratory overview](https://rail.eecs.berkeley.edu/). Accessed October 13, 2026.
13. UC Berkeley InterACT Lab. [Laboratory overview](https://interact.berkeley.edu/). Accessed October 13, 2026.
14. Tony Z. Zhao et al. ["Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"](https://tonyzhaozh.github.io/aloha/). Robotics: Science and Systems, 2023; [arXiv:2304.13705](https://arxiv.org/abs/2304.13705).
15. Moo Jin Kim et al. ["OpenVLA: An Open-Source Vision-Language-Action Model"](https://openvla.github.io/). 2024; [arXiv:2406.09246](https://arxiv.org/abs/2406.09246).
16. Alexander Khazatsky et al. ["DROID: A Large-Scale In-the-Wild Robot Manipulation Dataset"](https://droid-dataset.github.io/). 2024; [arXiv:2403.12945](https://arxiv.org/abs/2403.12945), revised April 22, 2025.
17. Physical Intelligence. ["openpi"](https://github.com/Physical-Intelligence/openpi). Repository documentation, accessed October 13, 2026.
18. NVIDIA. ["AI Robotics Platform, Tools & Simulation"](https://www.nvidia.com/en-us/industries/robotics/). Accessed October 13, 2026.
19. Tesla. ["AI & Robotics"](https://www.tesla.com/AI). Accessed October 13, 2026.
20. Stanford Report. ["Meet the robots at the new Stanford Robotics Center"](https://news.stanford.edu/stories/2024/11/meet-few-roboticists-designing-tomorrows-technology-new-robotics-center). November 19, 2024.