Sahith Chada

[Irene Grace @irenekarrot]

Got into @ycombinator for deploying robots into the real world

What do you think are the best usecases for generalist robots today?

[Sahith Chada @sahithchada]

Reach out if you need manipulation tasks that need automation!

[Irene Grace @irenekarrot]

Got into @ycombinator for deploying robots into the real world

What do you think are the best usecases for generalist robots today?

[Sahith Chada @sahithchada]

At #cvpr2026 in Denver this week! Excited to be here. Lots on my mind around VLAs and world models. If you're working in this space, come find me or shoot me a DM 👋

#Robotics #VLA #WorldModels #PhysicalAI

[Irene Grace @irenekarrot]

Got the i2RT YAM bot running with MolmoAct2 today.

Off-the-shelf capabilities continue to impress, even in terrible 2am lighting.

Prompt: "place all objects in container"

Waiting on leader arm parts to try fine-tuning

@allen_ai @DJiafei @i2rt_robotics

[Sahith Chada @sahithchada]

@lukas_m_ziegler Deploying generalist robots to warehouses

[Irene Grace @irenekarrot]

sold a kidney, got a YAM 🫡

Did a bunch of research and @i2rt_robotics have the best bi-manual bots out there. Can’t wait to set it up over the weekend.

[Sahith Chada @sahithchada]

@jyothi_kasina @irenekarrot @allen_ai @hq_fang @DJiafei Yep, ofc @allen_ai has trained the model on public so101 datasets, unlocking the zeroshot potential

[Sahith Chada @sahithchada]

@pham_blnh @irenekarrot @allen_ai @hq_fang @DJiafei Yeah, it works on a GPU with 16gigs memory using bf16

[Sahith Chada @sahithchada]

Being able to zero-shot policies on SO101 is a huge win for the open-source robotics community.
@huggingface @seeedstudio

[Irene Grace @irenekarrot]

Tried MolmoAct2 zero shot on SO101. genuinely impressed, can't wait to try this on a YAM bot.

Pick precision should be a lot better with the YAM model trained on depth, plus our wrist cam is literally a webcam right now lol.

Hats off to the team @allen_ai @hq_fang @DJiafei

[Irene Grace @irenekarrot]

Tried MolmoAct2 zero shot on SO101. genuinely impressed, can't wait to try this on a YAM bot.

Pick precision should be a lot better with the YAM model trained on depth, plus our wrist cam is literally a webcam right now lol.

Hats off to the team @allen_ai @hq_fang @DJiafei

[Irene Grace @irenekarrot]

Open weights + open data is how robotics actually moves forward. More of this please...

[Aryan Mangal @aryanmangal769]

RoboHacks @ycombinator ! Watch Rocky execute a long-horizon sequence of tasks in a single run, driven solely by voice command.
Team Rocky: @IndraneelPati10 @sakshamKuk70566 @sahithchada @a_andlyu @irenekarrot . Thanks @innate_bot @ax_pey @GoogleDeepMind @nasa @iterate!

[Alex Hovakimyan @alex8hovakimyan]

We made Jarvis, a robot assistant built for the 2026 Seeed Embodied AI Hackathon in sc.

Huge thanks to the amazing sponsors and ecosystem partners that made this possible: @seeedstudio @NVIDIA @PollenRobotics @huggingface @AgoraIO!

1/3

[Sahith Chada @sahithchada]

@eurofounder This shows the complete picture

[Sahith Chada @sahithchada]

@ankurnagpal Interesting

[Sahith Chada @sahithchada]

@shuman_p @CVPR I am unable to update the abstract

[Sahith Chada @sahithchada]

@stash_pomichter @dimensionalos Looking forward

[Sahith Chada @sahithchada]

@AlbertBoyangLi Aligning business interests with fundamental research is a big challenge

[Sahith Chada @sahithchada]

@iwantMBAm4 Honestly, DeepSeek may be under the radar here in the US, but in China, it’s anything but gone. It only cost $294K to train its R1 model, and deeper in China it's already driving gov’t deployments & beating rivals in benchmarks.