Digit Gets Smarter: How Agility Robotics Is Leveling Up with NVIDIA

Digit Is Already on the Job—And Getting Smarter by the Day

At NVIDIA GTC 2025, I had the chance to meet Digit—the humanoid robot from Agility Robotics—up close and in action. Digit is tall (5'9"), agile, and surprisingly commanding in person. As it carefully selected grocery items and dropped them into a bright red shopping basket (an odd juxtaposition of “having fun on your Target run,”) it was clear this wasn’t just a cute demo. (If you’re wondering, Digit decided upon long noodles and SPAM.)

In the attached video, the CTO of Agility Robotics, Pras Velagapudi explains more about Digit. This robot is the real deal: autonomous, adaptable, and already deployed in real warehouses, solving problems that have long resisted automation. It’s a working-class humanoid. And with NVIDIA’s simulation and computing platforms onboard, it’s learning fast.

Digit carefully selects novel items off the shelf.

Agility is now using NVIDIA’s Isaac Sim and Isaac Lab to train and test Digit’s behaviors through reinforcement learning—meaning this robot gets smarter by the (simulated) second.

“DIGIT BY THE NUMBER.” Image courtesy of Agiliry Robotics.com/solution.

But it goes beyond training. Using NVIDIA’s Omniverse “Mega” Blueprint, Agility can create digital twins of warehouses and factories—replicas of customer environments—where they optimize Digit’s performance before it ever sets foot on-site. It’s AI-enhanced robotics with a side of predictive logistics, and it’s already in use.

Read the full announcement from Agility Robotics released at NVIDIA GTC.

Also, a bit of fun trivia, Digit was one of TIME Magazine’s “Best Inventions of 2024.” Go, Digit!

#NVIDIAGTC2025 #AgilityRobotics #DigitRobot #HumanoidRobots #PhysicalAI #DeepLearningwiththeWolf #Robotics #ReinforcementLearning #DigitalTwins #NVIDIAOmniverse