XDOF
About
About XDOF
XDOF operates robot data collection at industrial scale, combining teleoperation hardware, trained operators, quality control and data infrastructure. Its systems capture synchronised camera streams and robot telemetry from bimanual manipulation sessions and organise the resulting episodes for model training and evaluation.
ABC-130K contains more than 130,000 teleoperated episodes, roughly 3,500 hours of data and nearly 200 tasks collected on two-arm YAM stations. XDOF released the dataset alongside hardware designs, training infrastructure and simulation tools, giving researchers an open dataset for robot-learning research.
Articles
Funding
- 2026-06-17 - $70M - XDOF raises $70M for robot data infrastructure
XDOF raised $70 million from Thrive Capital, Spark Capital, a16z, Lux and WndrCo to build data pipelines, collection tools and annotation systems for robotics teams. - source: techcrunch.com
Investors
- a16z - Participant - VC - 1 funding event - first seen 2026-06-17
- Lux - Participant - Unknown - 1 funding event - first seen 2026-06-17
- Spark Capital - Participant - VC - 1 funding event - first seen 2026-06-17
- Thrive Capital - Participant - VC - 1 funding event - first seen 2026-06-17
- WndrCo - Participant - Unknown - 1 funding event - first seen 2026-06-17
Team
- Philipp Wu - Co-founder and Chief Executive Officer - His research background connects XDOF's data infrastructure to teleoperation, robot-learning datasets and foundation-model training workflows.
- Fred Shentu - Co-founder and Chief Technology Officer - Teleoperation and robot-learning research form the technical basis for XDOF's data-collection systems for robot foundation-model development.
- Nemo Jin - Co-founder and Chief Operating Officer - Operations leadership for XDOF's data-collection, tooling and annotation business for robotics-model teams.
Research
- XDOF and research partners release ABC-130K teleoperation dataset
2026-06-17 - Research Publication - ABC-130K covers 134,806 episodes, 195 tasks and 3,553 hours of bimanual manipulation data. The ABC release includes open-source training and inference code, model checkpoints, a hardware setup and a simulation pipeline. The dataset is hosted on Hugging Face under the XDOF organization and is released under the Apache 2.0 license.
Source
Timeline
Events
- 2026-09-04 - Market Signal - XDOF reportedly enters Series B talks at $1.2B valuation
The first publication date was September 4, not September 7. Annualized revenue was approaching $50 million following XDOF's June Series A. No completed financing or confirmed valuation is recorded from the reported talks.
Source - 2026-06-17 - Research Publication - XDOF and research partners release ABC-130K teleoperation dataset
ABC-130K covers 134,806 episodes, 195 tasks and 3,553 hours of bimanual manipulation data. The ABC release includes open-source training and inference code, model checkpoints, a hardware setup and a simulation pipeline. The dataset is hosted on Hugging Face under the XDOF organization and is released under the Apache 2.0 license.
Source - 2026-06-17 - Funding - XDOF raises $70M for robot data infrastructure
The capital supports infrastructure for the robot-training data feedback loop, including data pipelines, collection tools and annotation systems. XDOF works with about 20 customers, including several frontier AI labs. XDOF launched in October 2024 to provide a data ecosystem for companies pursuing robotics models.
Source
Context
Comparable profiles
- StretchSense
stack adjacency - Embodied Data Capture
- FrodoBots
stack adjacency - Embodied Data Capture
- Liqing Intelligence
stack adjacency - Embodied Data Capture
- Vision Lab
stack adjacency - Embodied Data Capture · Physical AI Data Infrastructure · Robot Learning Infrastructure
- Wuwen Zhike
stack adjacency - Physical AI Data Infrastructure · Robot Learning Infrastructure
- Physen AI
stack adjacency - Embodied Data Capture · Physical AI Data Infrastructure
- Lightwheel
stack adjacency - Physical AI Data Infrastructure
- LabyrinthAI
stack adjacency - Physical AI Data Infrastructure
- MANUS
stack adjacency - Embodied Data Capture · Physical AI Data Infrastructure
- Rokoko
stack adjacency - Embodied Data Capture
- Maniformer
stack adjacency - Embodied Data Capture
- Jiiov / 极豪科技
stack adjacency - Embodied Data Capture
Shared workflows
- MOTOMIND
workflow adjacency - Teleoperation Data Collection
- Synthium
workflow adjacency - Robot Learning Data Collection
- Mind Robotics
workflow adjacency - Robot Learning Data Collection
- Hub
workflow adjacency - Robot Learning Data Collection
- TacnIQ.ai
workflow adjacency - Robot Learning Data Collection
- PHYMI
workflow adjacency - Robot Learning Data Collection
- Reward AI
workflow adjacency - Robot Learning Data Collection
- Moxian Technology
workflow adjacency - Robot Learning Data Collection
- Stouching Robotics / 视塔机器人
workflow adjacency - Robot Learning Data Collection
- AstrAlive
workflow adjacency - Robot Learning Data Collection
Context tags
- Teleoperation and data collection - Workflow - ABC-130K is a bimanual teleoperation dataset and XDOF traces its work to GELLO-style teleoperation systems.
- Robot-learning infrastructure stack - Stack Layer - XDOF describes its stack as data collection systems, operational capability, data warehouse and software tooling for robot learning.
- Physical AI data infrastructure - Stack Layer - XDOF builds robotics data pipelines, collection tools and annotation systems for frontier labs and robotics companies.
- Robot learning data collection - Workflow - ABC-130K and XDOF customer work center on teleoperation data collection for robot-policy training.
- Embodied data capture - Stack Layer - XDOF supports embodied-AI training data capture through robot demonstrations and physical-world interaction data.