Ashwin Balakrishna

@ashwinb96

Building robot brains @physical_int. Previously at @GoogleDeepMind, @berkeley_ai.

Bay Area, CA abalakrishna123.github.io Joined February 2011

231 Following 799 Followers 169 Posts

Latest posts 5

Pinned post

Ashwin Balakrishna @ashwinb96 Apr 16

I’ve seen dexterous robots before, and I’ve played with robots that showed some flickers of language following and generalization,
but I'd never seen a robot that’s really good at both! This new combination makes Pi-0.7 both highly capable and very fun to play with :)

Physical Intelligence Verified account @physical_int Apr 16

Our newest model, π0.7, has some interesting emergent capabilities: it can control a new robot to fold shirts for which we had no shirt folding data, figure out how to use an appliance with language-based coaching, and perform a wide range of dexterous tasks all in one model!

View quoted post on X

Ashwin Balakrishna @ashwinb96 Oct 13

Getting a robot to work in a sandbox is a demo. But being able to handle diverse hardware and tasks in the messiness of the real world is a platform. Super excited about this release!

Blog: pi.website/blog/api-beta

www.pi.website Frontier Models for Every Robot: Announcing our API beta Our latest robotics models are now available through the Pi API beta. Upload data, fine-tune models, and deploy cloud-hosted inference on your own robots.

Physical Intelligence Verified account @physical_int Oct 13

Announcing the beta release of the Pi API. We’ve partnered with companies across diverse industries to adapt our models to their robots and applications, from harvesting strawberries to assembling servers. Now we’re opening access to more teams.

View quoted post on X

Ashwin Balakrishna @ashwinb96 Aug 18

Excited for this fun chat about robot data!

Foxglove @foxglove Aug 13

5 days to #Actuate26! Down to our last few expo passes (will probably sell out by Saturday) so get your ticket today!

Excited to announce Tuesday's panel: Data Wars: Human Data vs UMI vs Teleop.

Moderated by @ryancjulian (@NVIDIARobotics), with Naveen Kuppuswamy (@FlexionAI), @philippswu (@xdofai), @ColinearDevin (@GeneralistAI), and @ashwinb96 (@physical_int)

Aug 18–19, SF Tickets: hubs.li/Q04sSWFQ0

View quoted post on X

Ashwin Balakrishna @ashwinb96 Dec 22, 2025

This was such a fun project to work on! The highlight for me: because of the strength of our pre-trained checkpoint, we didn’t need massive datasets. Just a few hundred demos were enough to get good policies for each of these super difficult tasks 🙂

Physical Intelligence Verified account @physical_int Dec 22, 2025

We got our robots to wash pans, clean windows, make peanut butter sandwiches, and more!

Fine-tuning our latest model enables all of these tasks, and this has interesting implications for robotics, Moravec's paradox, and the future of large models in embodied AI.

More below!

View quoted post on X

Ashwin Balakrishna @ashwinb96 Nov 18, 2025

A VLA that learns on the job 🙂

Sergey Levine Verified account @svlevine Nov 18, 2025

We just released results for our newest VLA from Physical Intelligence: π*0.6. This one is trained with RL, and it makes it quite a bit better: often doubles throughput, enables real-world tasks like folding real laundry and making espresso drinks at the office.

View quoted post on X