Full of childlike wonder. Teaching robots manners. RL Lead @ Apptronik. UT Austin PhD candidate. Past: Boston Dynamics AI Institute, NASA JPL, MIT ‘20.
he/him
Joined September 2018
- Tweets37.5K
- Following5.3K
- Followers16.4K
- Likes206K
Kyle🤖🚀🦭 retweeted
After we released the original Isaac Gym and then the Orbit framework, I took a step back from robotics to focus on 3D reconstruction and rendering - NeRFs and 3D Gaussians.
Then, in October 2024, I got a call from @JensenHuang , asking me to come back and take charge of Isaac Lab.
I was really enjoying being deep in the weeds working on technologies that were defining the future of 3D graphics. I also understood exactly what it would mean to take this on, and the level of commitment it would require.
So, I hesitated.
“It’s very flattering that you think I’m the right person for this, but I’ve been away from robotics for a couple of years, and things move fast. I don’t know our current strategy, or even who’s on the team now.”
Jensen’s response was simple, direct, and effective: “You decide.”
With that level of trust, the opportunity was hard to pass up.
Over the past two years, I’ve had the privilege of working with an extraordinary group of collaborators to define the future of robotics simulation with Newton and Isaac Lab 3.0.
I couldn’t be more proud of our incredible, hardworking team and what we’ve been able to accomplish together. Seeing everything come to fruition at GTC last week was amazing.
Thanks, Jensen!
We’re excited to announce the official release of NVIDIA Isaac Lab 3.0. 🎉
Isaac Lab 3.0 brings a new modular and lightweight architecture along with major new capabilities across the robot learning workflow: multi-physics and multi-rendering support, training at scale, and streamlined deployment on real hardware.
With Newton integration, the new OVRTX renderer, RL post-training, and LEAPP deployment, there’s a lot to explore.
Here’s a look at what’s new 🧵
(1/7)
the pace of research now is wild. supermimic came out 3 days ago. no code release. now there are already derivative works with full code.
it’s hard not to feel the pressure to always move faster.
One world, one policy.
MorphMimic shares a morphology-conditioned world model and policy across #Unitree G1, #AgiBot X2 Ultra, and #Booster T2—from the same dance in MuJoCo to a live browser demo and Unitree G1 hardware, with more to come. Code and report at lzyang2000.github.io/morphmi…
Huge thanks to the SuperMimic team for the inspiration!
Kyle🤖🚀🦭 retweeted
We present OpenTac, an open, full-stack ecosystem for 3D tactile sensing on any robot surface.
Anyone can design and fabricate high-density tactile skins for any part of their robot, then read them out, simulate them, and share data across different sensors.
Kyle🤖🚀🦭 retweeted
Hello again, it's been a while.
We're writing to let you know that we have raised quite a lot more money. We're letting you know because we are on the lookout to hire smart people and this will help us do that.
See you again when we have a product or have raised more money!
I genuinely do not understand why this is the demo they chose to release
unrelated, how long has the word “civilian” been in their bio?
Kyle🤖🚀🦭 retweeted
What frustrates me in some “agentic robot planning” work from the ICLR submission is how little it engages with either classical planning algorithms or the lessons of scaling. We shouldn’t build bespoke architectures as though history doesn't exist, then treat a SOTA number as sufficient justification. (1/N)
can we stop with the LinkedIn slop paper hype :/
I don't think people realize how insane @younghyo_park and the folks at @pulkitology's lab are. Incredible results, psycho execution strategy.
1. They managed to fit pegs and gears with joint states sensing only - so it can work on the most basic $2k QDD arm, no camera, no tactile sensing on hands.
but how
2. No simulator is good enough to train for this - so they built their own that's precise enough for tight fits/screws/gear meshing called mjfit
but it doesn't learn in RL + standard network architecture, so
3. Bag of tricks were experimented on and implemented:
a. Priviledged sensor data for training critic
b. Temporal convolution before passing into MLP
c. Parameterization of parts that allows for insane amounts of randomization
but is that really why it works
4. Analysis experiments on the learnings + visualizations, simplifying to
a. Blind, but not open-loop
b. Visually blind, yet intelligent
c. Proprioception generalizes better
5. This generalizes into the real world
Beautiful papers like this exist, where people have put in exceptional effort and then decide that its value is best exemplified as open research. Now somebody is going to use this to post-train a VLA, or to distill a world model from simulation.
Thanks brother, the shoulders of giants we stand on grows another feet taller.
younghyopark.me/blind-propri…
Kyle🤖🚀🦭 retweeted
Manipulation gets solved by feeling; vision isn't as important as you think! A leap in the last mm manipulation that holds back general-purpose robotics 👇
Every robot already carries contact sensors: its own joints. When a robot touches something, its joints drift slightly off their commands, and the encoders pick it up. Legged robots and humanoids already leverage this signal to walk blind - but is it enough for blind precise assembly?
We find that it is! Our policies can insert pegs, mesh gears, and thread nuts using only the robot's joint positions - no camera, no tactile, and no object poses. And the peg policy works zero-shot in the real world! 🧵
w/ my amazing co-first author @younghyo_park, and advised by @kushalk_ and @pulkitology.
I am genuinely not joking when I say: a Mac mini
Hate to dox myself a bit but have worked on just about every fighter jet the US has. I have seen some impressive technology over my career but this is a whole other level. I have so many questions and need to know how this works. How much edge compute is needed to make this happen? My first thought is dual processor + 1TB ram. That way it can handle the fuckquintillion process per second. There was zero correction it nailed it 100%.
Kyle🤖🚀🦭 retweeted
I really like these guys, I think it's the best open-sourced hardware out there
🚀RP1 is officially here.
🌍Putting control of robotics back in developers’ hands.
Introducing RP1 (ROBOTO 01) by RoboParty—a high-performance, fully open-source humanoid robotics platform for frontier robotics research and embodied AI model development.
🤖 Open across the full hardware and software stack:
• Robot platform: mechanical CAD, bill of materials, and assembly documentation
• Actuators: in-house ROMOMO joint designs and related technical resources
• Control: electrical systems, firmware, low-level drivers, and motion control implementations
• Development: simulation models, SDKs, data processing tools, and training tools
• Engineering foundations: technical documentation and research materials
Open the black box.
Understand the system.
Make it your own.
A deviation in motion can originate in the model, control policy, actuator response, or mechanical structure.
RP1 opens up the design and implementation of these layers so developers can trace problems, modify the system, and validate changes on real hardware.
Open-source algorithms for advanced motion control
RP1 has demonstrated continuous walking, recovery from disturbances, getting back up after a fall, and human motion tracking.
UFO and MimicLite connect motion learning, control policies, and execution on real hardware.
The open toolchain connects data collection, simulation training, hardware deployment, and experimental validation—supporting joint iteration across models, control, and hardware.
From RPO to RP1, we’re turning one team’s engineering experience into a starting point for many more.
Opening the future of embodied AI to everyone ready to explore it.
🔗RP1 on GitHub:
github.com/Roboparty/RP1
#RoboParty #RP1 #OpenSourceRobotics #EmbodiedAI
Kyle🤖🚀🦭 retweeted
This is some super solid work led by @jakejlevy at UT and Tyler Westenbroek from UW earlier this year! Re-reading this one because it honestly feels like it should be a baseline for a subset of the DARPA TIAMAT challenges. Excited to see what's getting cooked up next!
I’m presenting SimDist at #RSS2026 today! This framework takes failing policies to ~90% success with only 15-30 minutes of data. Meet me at the poster or reach out to chat!
💬 Presentation: “World Models & Memory”, 3:15-4:00PM
📜 Poster: 6:30-7:30PM
🌐 sim-dist.github.io/
this is pretty wild for a DVA that was likely trained on many orders of magnitude more video data than tactile data
