Skild AI needed a way to train robots that can adapt across many bodies, tasks, and environments, but collecting enough real-world robotics data was too slow and expensive to produce robust general-purpose physical AI.
Reported impact
Skild AI achieved 60-80 percent task performance within hours of data collection, enabled robots to carry payloads up to 1.5x body weight, reduced total cost of ownership by 10x compared with traditional systems, and demonstrated recovery from jams and limb failures through in-context adaptation rather than reprogramming.
Skild AI built the Skild Brain as an omni-bodied robot foundation model on NVIDIA accelerated computing and simulation tools so it could train physical AI without depending on massive real-world robot fleets. The company uses NVIDIA Omniverse libraries and NVIDIA Isaac Lab to create large-scale physics simulations across thousands of robot embodiments and environments, while NVIDIA Cosmos Transfer augments those datasets with varied lighting and visual conditions. This synthetic pipeline is paired with human videos from the internet to teach robots how objects afford actions, giving the model broader behavioral coverage than teleoperation data alone. In deployment, the Skild Brain consumes raw camera images and proprioceptive signals and directly outputs low-level motor commands, enabling zero- or minimal-post-training adaptation to new robots. The result is a robotics stack that learns through scalable simulation and data augmentation rather than narrowly programmed behaviors.
Synthetic robot scenarios and internet-scale human videos are fed into Skild AI's training pipeline, Isaac Lab and Omniverse generate scalable simulation experience, Cosmos Transfer augments the data, and the resulting Skild Brain turns vision and proprioception into low-level motor commands for new robot bodies in real-world tasks.
"NVIDIA Isaac Lab and Cosmos technologies allow us to create massive and scalable data sources necessary for robots to truly learn from experience across diverse scenarios and embodiments."
Skild AI built a true robotics foundation model called the Skild Brain.
The company uses NVIDIA Omniverse™ libraries and open frameworks such as NVIDIA Isaac™ Lab for advanced physics simulation and NVIDIA Cosmos™ for data augmentation and generation to train its foundation model.
The model demonstrates remarkable adaptability to mechanical changes, recovering from jammed wheels within 2-3 seconds.
This multi-pronged simulation approach enables Skild AI to acquire a millennium of experience within days.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 15, 2026 Lambda Lambda needed to raise token throughput and improve AI-cloud economics, but fixed rack and facility power budgets were leaving GPU capacity stranded and limiting how many nodes could run safely at once. +17% Hybrid Training Throughput Dynamic Power Software · NVIDIA DSX AI Factory Platform · NVIDIA DSX MaxLPS NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph 