Senior Software Engineer, Bloomberg Hobby robotics in the garage. Sonic X2 Port - https://xitter.cf/t.co/NKdbcLvrlx

California, USA
Joined September 2009
Spent months training X2 Sonic from scratch. A frozen G1 model just beat it. 😳🤯 The ~40M-parameter model itself untouched — just a joint codec + LoRA adapters, trained overnight at ~2% of its original compute. sonic-agibot-x2.github.io/so…
6
25
2
170
13,548
SuperMimic is now running on the @AGIBOTofficial X2 Compared with SONIC on the same X2: In-distribution success: 98.4% vs 98.4% Out-of-distribution: 82.6% vs 84.0% Tracking error (MPJPE): 19 mm vs 30 mm Trained 36 hrs on a single H100 github.com/meetsitaram/GR00T… Dancing Gangnam from Boneseed.
We found a way to scale humanoid control that reduces GPU-hours by more than 1000×. Across 129,514 motions, SuperMimic surpasses the state of the art (128 GPUs, 7 days) with 1 GPU in 3 hours. Project: supermimic.github.io Demo: supermimic.github.io/demo/
4
4
2
45
2,907
It still has some issues with normal walking, when combined with motionbricks kplanner, that need to be ironed out.
2
133
Great times ahead.
We found a way to scale humanoid control that reduces GPU-hours by more than 1000×. Across 129,514 motions, SuperMimic surpasses the state of the art (128 GPUs, 7 days) with 1 GPU in 3 hours. Project: supermimic.github.io Demo: supermimic.github.io/demo/
2
255
Your mission, should you choose to accept it: rescue the robots from decommissioning by training the ultimate motion control and planner models, and set them free. 🤖 github.com/meetsitaram/robo-… Sonic and kplanner models by Nvidia; Motion data by Bones Studio; Video by Grok and Claude; Music: Transformers Uprising by LumineWave
9
513
Exoskeleton - Fusion360 CAD designing with just prompts - github.com/meetsitaram/cdx-3…
1
2
11
1,377
Sitarama Chekuri retweeted
Starship transiting the Sun this morning during Flight 14 — prior to this morning's mission, this has never been captured before with the world's largest and most powerful rocket. I’ve ached over this shot for years, mainly for two reasons: what does methane exhaust look like silhouetted against the Sun, and how violent are the acoustic energy waves from 33 Raptor engines on the Super Heavy booster? Well...now we have an answer. 📸 - @NASASpaceflight
474
1,854
225
12,277
977,576
Lock In
Gave Opus 5.5 donald's prompt, Midjourney, and a moodboard 12 hours later, woke up to this:
143
Sitarama Chekuri retweeted
hilarious to watch how Opus 5.5 designs a 3d asset, prints it, and controls a robo arm to take it out of the printer
56
94
37
1,522
129,263
Sitarama Chekuri retweeted
How do you teach a humanoid to assist another person in close-contact? 🤖 The hard part: the two bodies are physically coupled — helper & helped continuously shape each other's motion. Neither can be solved alone. Meet AssistMimic, our multi-agent RL framework👇🧵 #CVPR2026
4
32
3
165
18,432
No IMU? No problem. Welcome to the world of Blind Dexterity
In my latest PhD paper, we declare WAR on sensor-maxxing. For the first time, push-resilient humanoid walking, just with joint encoders! No IMU, no F/T sensors. 🥁 Introducing Blind Dexterity 🧵 👇 w/ @OKaidanov @liu_puze @Jan_R_Peters at @DFKI @ias_tudarmstadt
10
1,179
Sonic on the big guy 😍
SONIC support for Agibot A3/A3 Ultra is now open source. Includes A3 training code, MuJoCo sim2sim, C++ deployment, PT/ONNX/RKNN weights, and RK3588 NPU inference. Also compatible with A3 Ultra. Huge thanks to our team and NVIDIA GEAR. 🔗 Repo: github.com/AgibotTech/sonic_… 🎬 Demo: agibottech.github.io/sonic_f…
5
623
Sitarama Chekuri retweeted
In 12 weeks, we built a research facility that is run entirely by AI. AI designs, executes, and observes experiments end-to-end across biology, chemistry, and materials science. We’re introducing SciUniverse: a benchmark that measures AI’s ability to do real-world scientific research.
243
394
220
2,818
804,963
Sitarama Chekuri retweeted
Fast dribbling. Tight turns. Repeated direction changes. Human-like motion. ⚽️ We present a Unitree G1 system for sustained humanoid dribbling beyond simple straight-line ball pushing. The robot continuously controls the ball while running through slalom courses and large direction changes, using natural whole-body motions. Ball perception is fully onboard with LiDAR — no mocap, no external tracking. The policy is trained in simulation with a human-motion prior and transfers zero-shot to hardware. Safety tether stays slack during operation and provides no assistance. Paper coming soon. Stay tuned!
2
17
10
108
49,094
Self-Play training with @nvidia Gear Sonic and @TheBonesStudio combat moves for the Ultimate Bots Season 2
ULTIMATE BOTS SEASON 2 IS HERE. 🤖🏆 4 teams. 6 events. $100K on the line. Teams will earn points across 3 disciplines: 🥊 Martial arts 💃 Dancing 🏁 Obstacle course Season 2 hits F1 Austin → TwitchCon → IEEE Humanoids → CES → Big Game Weekend → NVIDIA GTC. Captains + teams announced soon. Follow for reveals, ticket drops & live streams.
1
26
2,945
Obstacle Course Track 🔥
ULTIMATE BOTS SEASON 2 IS HERE. 🤖🏆 4 teams. 6 events. $100K on the line. Teams will earn points across 3 disciplines: 🥊 Martial arts 💃 Dancing 🏁 Obstacle course Season 2 hits F1 Austin → TwitchCon → IEEE Humanoids → CES → Big Game Weekend → NVIDIA GTC. Captains + teams announced soon. Follow for reveals, ticket drops & live streams.
2
194
Out for delivery
10
689
Magical
Replying to @Figure_robot
Here's 4 hours of our robot doing zero-shot work in 30 rental homes
140
Sitarama Chekuri retweeted
A video from a Pi robot deployed at Dandelion Chocolate, fully autonomous w/ no interventions. 🤖 Deploying robots has taught us surprising lessons about the gap between proof-of-concept (i.e. building one box) and real-world utility (productively building boxes for hours). I expected that the hard part is building the box, since it’s the most dexterous, but that wasn't what we found. Counterintuitively, the hardest part was reliably stacking the boxes. Our original table-mounted robot had poor visibility of the stack without special separately-mounted cameras. Plus, stacking requires more generalization (each box is placed in a different location), and an imprecisely-placed box can lead the entire stack to collapse many boxes later. We recently switched this deployment over to a mobile robot, and it's fun to watch the robot being completely self sufficient for multiple hours. 🙂 Data and feedback from real world use-cases like these are quite valuable for the π pre-trained model as we scale!
46
67
27
923
105,956
1x speed! That is cool.
Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids. - learned directly from human manipulation data - no teleop/robot data - close to human-level dexterity and efficiency - multi-robot collab
9
1,178
Sitarama Chekuri retweeted
Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids. - learned directly from human manipulation data - no teleop/robot data - close to human-level dexterity and efficiency - multi-robot collab
154
478
291
3,498
1,774,011