A startup called XDOF emerged from stealth in June 2026 with $70 million in funding and a mission to solve one of physical AI’s most pressing problems: the lack of high-quality robot training data.
The San Francisco-based company, founded in October 2024, has raised its seed round from Thrive Capital, Spark Capital, Andreessen Horowitz, Lux Capital, and WndrCo. It currently employs around 60 people and says it is already working with 20 customers, including several frontier AI labs it declined to name.
XDOF was co-founded by CEO Philipp Wu, CTO Fred Shentu, and COO Nemo Jin. Wu previously studied at UC Berkeley, where his robotics research was stalled by the very problem XDOF now addresses. “We didn’t have large-scale data to work with,” he told TechCrunch. “There was this chicken-and-egg problem — we first needed to actually collect data before we could even ask how to train a foundation model for robotics.” Wu and Shentu previously collaborated on GELLO, a low-cost teleoperation system for generating robot training data that became widely used in the field.
Unlike language models, which were trained on vast amounts of publicly available text, robots require data that captures physical interaction — footage that barely exists at scale and is difficult to use in its raw form. XDOF plans to build data pipelines, collection tools, and annotation systems across three tiers: teleoperation data collected on deployed robots, more general teleoperated robot data, and “egocentric” data gathered by humans wearing sensors XDOF intends to build itself.
As part of its launch, XDOF is partnering with UC Berkeley’s AI Research lab to release a dataset called ABC, which includes 130,000 trajectories of robot manipulation data, 300 hours of simulation, and 100 hours of evaluations — described as the largest collection of high-quality robot training data ever assembled for academic use.
The company’s name is a play on “degrees of freedom,” a robotics term describing the number of independent motions a robot can perform. Wu said the X represents “arbitrary degrees of freedom, unlimited degrees of freedom” — reflecting the scope of physical tasks the company’s data is intended to support.
The launch comes as major AI labs, including OpenAI — which announced in 2026 it would relaunch its robotics program after shutting it down in 2021 — are racing to develop machines capable of operating in the physical world. Wu said the urgency is real: “You don’t want to be in this type of situation where you pursue this technology too late.”
Source: TechCrunch