NVIDIA NVIDIA
April 22, 2026

One Brain, Any Robot: Skild AI's Skild Brain Explained | NVIDIA AI Podcast Ep. 295

What if one AI brain could run every robot on the planet—a humanoid, a warehouse arm, and a dog-like inspection bot—all at once?

That's not a thought experiment. That's what Skild AI is building...

What if one AI brain could run every robot on the planet—a humanoid, a warehouse arm, and a dog-like inspection bot—all at once?

That's not a thought experiment. That's what Skild AI is building right now.

Deepak Pathak (CEO and Co-Founder) and Abhinav Gupta (President and Co-Founder) of Skild AI break down Skild Brain—a universal, general-purpose AI model designed to power robots of any form factor, tackling any task, from a single shared intelligence.

🔬 Topics covered

Why robotics is fundamentally a data problem—and why there's no "internet of robot data"

Skild's three-layer data strategy: real robot teleoperation, video pretraining, and simulation

How the pretraining/post-training recipe that made LLMs work is now being applied to physical AI

The data flywheel—how every robot deployment makes the shared brain smarter for the next task

The deployment roadmap: factories and warehouses → hospitals and hotels → home consumer robots

Skild's NVIDIA partnership: co-developing Newton physics solvers and using Cosmos for data augmentation, Isaac Sim for simulation, and edge compute for on-device robot inference

How Skild tests Skild Brain before deployment—from task KPIs to generalization stress tests to safety guardrails

When will robots actually be ready for your home? The co-founders give their honest (and surprisingly humble) take

Resources mentioned

Skild AI

Skild AI Blog

Chapters

00:00 Introduction: What is Skild AI?
03:19 Why robotics needs a horizontal platform
03:50 The 90% wall: how corner cases killed traditional robotics
08:34 The three sources of robot training data
10:57 Pretraining vs. post-training in physical AI
16:32 The data flywheel and cross-vertical deployment strategy
17:34 How Skild AI works with NVIDIA (Isaac Sim, Cosmos, Newton, edge compute)
18:59 Testing Skild Brain: task KPIs, generalization, safety guardrails
23:06 The future of robotics: factories now, homes later?
27:54 What's next for Skild AI
535

Comments 0

Replying to
5000 max

Be the first to comment!

This website uses cookies
This website uses cookies.
To maintain our services, we use cookies. They allow us to personalize ads, analyze our traffic, and provide social media features.

Customize your preferences

Select the cookies you accept: