Director, AI & Robot Behavior, Atlas @BostonDynamics Previous life: Faculty @MIT, PhD @CMU_Robotics
Joined May 2026
- Tweets27
- Following71
- Followers1.9K
- Likes26
You know there's a lot of robotics hype when a friend sends a photo of you on a billboard! I'm excited to catch up with @Ken_Goldberg , @parada_car88104, @deepakpathak and many friends at Fort Mason. See you all soon!
Today we give Atlas dexterous hands, with the strength to take on real work, and the simplicity for mass manufacturing. Like the body, the hands are designed with high fidelity simulation in mind, to enable sim-to-real RL. If you give dexterous hands to a robot, make them do dexterous things.
More thoughts here
Blog: bostondynamics.com/blog/robo…
Video: youtube.com/watch?v=4whgw2gL…
Once your data ops are humming, the real bottleneck is diverse, application relevant manipulation data. Data farms are too often focused on volume.
RMAC puts our data flywheel inside one of the richest sources of novel manipulation scenarios in the world. Spend enough time on a factory floor and you realize: manufacturing is practically manipulation-complete.
youtube.com/watch?v=CFS_zU1k…
Sometimes a writer digs past the immediate campaign to piece together the longer-term story. @johnkoetsier at @Forbes did a great job at that for the Atlas at the WorldCup campaign: hardware simplicity, behavior generality, and deployment strategy.
forbes.com/sites/johnkoetsie…
Want to row with this crew? Careers page link is in the comments.
A special callout for openings in our whole body and manipulation RL teams of senior positions for researchers with several years of hands-on experience. If you want to work in RL on the most capable humanoid alongside an incredible team, feel free to message me directly.
I just read @saranormous recent “The Untrainable,” and it resonates with how I see the next 5-10 years of robotics and humanoids. When I hear humanoid deployment timelines scaling to hundreds of thousands of units “in the next couple of years”, I think those figures come from a deliberate choice to ignore the untrainable realities of scaling a robotics product.
General robot behavior models are key to building a viable humanoid product, and the models are moving fast. But the product also scales on many other factors that simply do not move at the pace of the model:
- Data acquisition at the edge, where you have to work closely with customers to access their proprietary, high-value tasks and translate them into quality data to feed the model.
- Supply chain maturity, to secure a hardened supply chain that enables manufacturing at mass scale with high reliability.
- Fielding the product, establishing field deployment infrastructure, robust customer service loops, and kick-starting the data flywheel that improves the model with in-distribution data while preventing behavior degradation.
All of these are won the hard way, and they sit squarely in the untrainable corner. Humanoids will get there, but there is no way to separate the humanoid from the humanoid product.
What a ride bringing Atlas to the @FIFAWorldCup, from a dramatic entrance to delivering the official match ball to the referee.
Yet another great collaboration between @BostonDynamics & Hyundai Motor Group to showcase the now and tomorrow of robotics on the world stage.
Last week I sat down with @NBTJacklyn and Kanishka Rao from @GoogleDeepMind to talk about the state of robotics and AI at one of the Dialogues at Google I/O.
At some point the conversation touched on the two-system architectures we use to drive robots today. While not a perfect division of roles, one captures the more physical aspects of intelligence, like body control and interaction with the environment, while the other captures the more cognitive and reasoning parts of intelligence.
The act of sitting down is actually an interesting example of physical vs. cognitive. Around 6 to 9 months, children master the physics of sitting, they learn to control their torso, interact with surfaces, and maintain an upright back and head without falling. It isn't until much later, around a year and a half, that they begin developing the common sense of what constitutes a good place to sit, and creatively find novel fun places to sit on.
While eventually the physical and the cognitive complement each other in ways that are difficult to separate, I suspect that the order in which these two are learned matters. Mastering physical intelligence first must be key to developing an understanding of the world that is rooted in physical common sense later.
Thanks to @parada_car88104 and the I/O team for the invitation!
youtu.be/jn3iypY-cN4
Getting ready for @FIFAWorldCup !
Here is my take on one of the questions that I’ve heard most often recently: why humanoid?
The thing that drives the current bet on humanoids is the perceived potential value of a generalist solution to physical work. Outside of some applications that have very large and very stable volumes, for all other physical work, automation becomes unfeasibly expensive if we have to specialize hardware, or models or deployment strategies.
Something core to humanoid platforms today are also the strategic departures from the human form factor.
For example, we’ve moved entirely to rotary actuators, leaving behind linear actuators that more closely mimic the form factor of our musculo-skeletal system, because rotary actuators are more efficient and have a smaller sim2real gap. This has negative implications, for example for distal mass at the ankles and wrists, but the benefits from focusing on designing just two highly efficient actuator types have brought important gains.
The actuator efficiency allows us to rely on implicit proprioception to adapt to unexpected loads, without any force/torque sensors. And the increased reliability and smaller sim2real gap has increased our development speed.
(you'll see at the end of the video an unexpected load doubling the weight of the fridge)
Another key departure from human morphology is the infinite rotation of the actuators by removing all cables across joints. Eliminating all cables across joints gets rid of one of the key hardware failures and gives Atlas its unique superhuman mobility. Atlas can move backwards simply by inverting its legs and flipping its torso, rather than wasting steps and time turning around.
One of my favorite things is seeing Atlas stand up. It makes it clear we have not built Atlas to be a human replica. To me it looks like it belongs in the Star Wars or Transformers universes. Meant to be intentional, effective, and useful.