Banseog note · 2026-10-26
A research company aiming to supply the brain for many robots
Current stage · from public events · 2026-11-14
- R&D · prototype
- Pilot · first production
- Scaling production
- Operations · field rollout
- Going global · localizing
- M&A · integration
- Restructuring
Pilot · first production Announced limited release of robot model training and inference APIs to a small number of companies and initial field deployments under way
- 2026-10-13 The company released a beta of its training and inference APIs and described early access for a small number of companies and initial deployments in Japan, Switzerland and the US. Source
Roles often seen at this stage Robot software, AI & vision · Controls & embedded · Field, integration & service
Companies at the same stage → · Same stage in other industries →
- Founded in 2024 by former Google DeepMind researchers and Stanford and UC Berkeley academics, the company is headquartered in the Mission District of San Francisco.
- In November 2025 it raised USD 600 million led by CapitalG, taking total funding past USD 1 billion (Lux Capital, Thrive Capital and Jeff Bezos reinvested).
- Its public models π0, π0-FAST and π0.5 were trained on more than 10,000 hours of robot data and are available as open source in openpi (Apache-2.0).
Decisions within a year
- Pi API beta on October 13, 2026 it opened the Pi API beta offering hosted fine-tuning and cloud inference, naming Ultra, Chef Robotics and Formic among its early users.
- π0.7 release on April 16, 2026 it announced π0.7, which combines learned skills to perform new tasks, and said a single model controls multiple robots, including mobile manipulators, dual UR5e arms and Franka arms.
- π*0.6 release in November 2025 it released π*0.6, which learns from experience through reinforcement learning, and said it more than doubled throughput over the base model.
People they will look for
- Robot learning researchers (VLA, reinforcement learning) — foundation model research such as π0.7 and π*0.6 is the core of the company
- Customer technical support (solutions) engineers — a new role helping outside robotics companies using the Pi API with data collection, fine-tuning and deployment
- Robot hardware and data operations staff — collecting training data across multiple robot types drives model performance, and the company says it also hires in hardware and operations
For hiring managers
In Korea, large companies' AI research labs, humanoid and robotics startups, and the Korean organizations of global big tech firms are looking for the same VLA and robot learning researchers. This company says it hires only a small number, but as more companies like it appear, talent will move more often, so it is important for Korean companies to hold on to their robot foundation model talent now.
For job seekers
Experience directly training and deploying VLA or reinforcement learning models on robot data is what this company values most right now. The work location is San Francisco, and hiring is small in number.
An analysis written from public sources. Sources: SiliconANGLE 2025-11-20 (USD 600 million raise, total over USD 1 billion, founding background, π*0.6), Physical Intelligence website, checked 2026-10 (research release schedule, Pi API), Physical Intelligence blog 2026-04-16 (π0.7), Unite.AI 2026-10-13 (Pi API beta, early users, training data for public models), Physical Intelligence careers page (Mission District, San Francisco; hiring areas).
Roles hiring now
Skills asked for most
Open jobs
Banseog checks official company careers pages every day. A job is marked closed only when the official page returns a removal (404 or 410).