Research Engineer @GoogleDeepMind; prev cofounder @omegalabsai and AI @WOMBO 🇨🇦/🇵🇰
Mountain View, CA
Joined April 2020
- Tweets465
- Following4K
- Followers1.5K
- Likes10.2K
Salman retweeted
Gemini 4 Argon is a step change and will have massive impact on robotics. We have published a developer blog to show various ways in which Argon accelerates robotics.
developers.googleblog.com/ge…
Salman retweeted
🤖🍨 Next Sundae (luma.com/l7zqfsl1) on Oct 18th, @jamesjingxi and I are excited to host @XiaomengXu11 (PhD @ Stanford REALab) for Robots That See and Act with Their Whole Body... from whole-body manipulation, active perception & hardware-software co-design to RoboPanoptes, PanoVine, and HoMMI, exploring how novel robot embodiments, distributed sensing, and learning can enable robots to see and act with their whole bodies.
Salman retweeted
The point of man is to build, create, and improve the lives of others.
The number on the screen is not relevant.
Salman retweeted
Can we build world models purely in the latent space, without pixel prediction?
We present Contrastive World Models - we train the latent states to maximize mutual information with future observations, without any decoders / pixel reconstruction.
Contrastive World Models enables
⚡ Substantially more robust representations
🤖 More efficient training by removing the pixel decoder entirely
🌍 A general approach with minimal assumptions [1/n]
Salman retweeted
A huge thank you to everyone who joined @SundaeRobotics 08! 🤖🍨 Special thanks to @tylerlum23 for an outstanding talk on Dexterous Manipulation, Simulation & Precise Assembly, from SimToolReal and zero-shot sim-to-real tool use to Play Pretraining, Precise Assembly, and Play2Perfect.
Next week, we’re heading to Pittsburgh for IROS and bringing together a small, curated group of researchers, founders & builders for dinner. If you’re attending IROS and working on something interesting, request an invite: luma.com/wchju499 🤖🍽️
Thanks @drjingxi @menemazarakis @Joshuabrowder
Salman retweeted
Building in the New Spatial Intelligence Era · @SpaceFronClub × @MissionRobotBay
On September 23, Spatial Frontier Club and Mission Robotics are once again bringing the community together in San Francisco.
🥂 We’re overwhelmed by the enthusiasm and support, with over 500 registrations and our event being listed again as featured in San Francisco by Luma. 🔥🔥
Our speakers:
🔹 Salman Shahid @_sshahid_ | @GoogleDeepMind
Research Engineer working on Embodied Reasoning, with a focus on training dexterous robot policies while preserving model generalization.
🔹 Hokin Deng @DengHokin | @CarnegieMellon
Researcher studying how AI understands and predicts the physical world from video, and lead of VBVR, a video model benchmark spanning 56 authors across 32 universities.
🔹 Jake Fitzgerald @earthtojake | @spc
Creator of an open-source project with 16K GitHub stars and 200K downloads. Jake will share how he used AI tools to build a six-degree-of-freedom robot arm without prior robot-building experience—and what he learned about using LLMs to build hardware.
🔹 Andrey Gizdov @andreygizdov | @OpenVectorLabs (YC S26)
CEO and Co-Founder of OpenVector, a YC-backed startup building Vision-Language-Action (VLA) systems to understand and track real-world processes. His research spans compute-efficient multimodal models and spatial reasoning, with publications at CVPR, ECCV, and AAAI.
Can't wait to meet everyone for an incredible evening!
📍 San Francisco
📅 Wednesday, September 23
🕠 6:00 - 9:00 PM
Register on Luma: luma.com/3kwuatyr
#SpatialIntelligence #WorldModels #Robotics #PhysicalAI #SpatialFrontierClub 🔥 🔥
Salman retweeted
A huge thank you to everyone who joined
@SundaeRobotics 06! 🤖🍨 Special thanks to @DanielDugas14 for an outstanding talk on V-JEPA 2 and predicting physical intelligence.
Next Sundae (luma.com/a0einzvt), we're excited to host @LinfengZhaoZLF (Postdoc at the Stanford Robotics Center, working with Prof. Kochenderfer and Prof. Jeannette Bohg) for a talk on Long-Horizon Robot Manipulation, Hierarchical Control & World Models with Retriever.
Thanks @menemazarakis @Joshuabrowder @drjingxi
@angelajiazhang
Salman retweeted
A huge thank you to everyone who joined @SundaeRobotics 05! 🤖🍨 Special thanks to @bardienus for an outstanding talk on 3D World Models, Spatial Intelligence & Adaptive AI.
Next Sundae (luma.com/5h9owhri), we're excited to host @DanielDugas14 (AI Researcher at Flexion Robotics, previously at Meta FAIR) for V-JEPA 2 & Predicting Physical Intelligence... from self-supervised world models and latent prediction to robot planning, value learning, and Physical AI post-training.
Thanks @PaulKefer @thejonohart @menemazarakis @Joshuabrowder @drjingxi @angelajiazhang @BlaineGame
Salman retweeted
today, we’re releasing the largest open-source human image preferences dataset, along with a $1 million data grant
- 2M+ annotations by real people
- 30 SOTA image models ranked
- 10 categories (marketing, product design, anime etc)
dataset + benchmark + grant details below:
Salman retweeted
A huge thank you to everyone who joined @SundaeRobotics 04! 🤖🍨 Special thanks to @kaiwynd for an outstanding talk on the future of Robotic Simulators.
Next Sundae (luma.com/f6t112qi), we're excited to host @BDuisterhof, a final-year Ph.D. student at CMU advised by Prof. Jeffrey Ichnowski, for a talk on 3D World Models, Spatial Intelligence & Adaptive AI with Modality Forcing & 3D Pre-training Objectives.
Thanks to @angelajiazhang @thejonohart @menemazarakis @drjingxi @Joshuabrowder
Salman retweeted
The Lab is open.
Interfaces, agents, simulations — specimens of the working web.
laboratory-mocha.vercel.app
Salman retweeted
Yesterday was my last day @spacexai.
I’m incredibly grateful to have had the opportunity to be a part of xAI’s journey in the last year.
I got the chance to work on 3 different teams, contribute to challenging problems, and hangout with cool folks that I now call great friends.
Special shout out to @VastoLorde95 and the entirety of the X engineering team for making my experience unforgettable. From changing my country’s flag 🇮🇷 (s/o @nikitabier 🐐) and rewriting the twitter ad stack, to f_score, and go/nads, I made so many memories that I’ll always cherish.
Thank you @elonmusk! I learned a lot from seeing you operate at trillion scale.
Much love,
Parsa
P.S. s/o @jon for getting the ads team the legendary mariachi band
P.P.S. First pic is the charitable view warroom. We pretty lived there for 12 weeks. The photo was taken on the last day of the migration (had to do 40h this time 💀)
P.P.P.S. I got the grok sticker on my laptop in my 2nd month in a recruiting event organized by @barisakis 🐐
P.P.P.P.S. Fun fact, I actually broke my foot in dec and was hopping around on crutches at office for a while. Good times
Salman retweeted
A huge thank you to everyone who joined @SundaeRobotics 02. 🤖🍨
Special thanks to @ChanghaoWang78 for an outstanding talk on the future of human demonstration, DexterityGen & SPIDER.
🤖🍨 Next Sundae (luma.com/th35npqz), we're excited to host @binghao_huang, a Ph.D. student at Columbia University advised by Yunzhu Li, for a talk on scaling tactile intelligence for robot learning and flexible tactile skin.
Thanks to @menemazarakis @angelajiazhang @thejonohart @drjingxi @Joshuabrowder!
Salman retweeted
One brain. For any robot. 🤖
We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
Salman retweeted
We just raised $7.2M USD to make insurance feel magical.
Nobody has ever said "my insurer was so easy to deal with."
We're changing that.
@generalmagic_ai builds AI agents that text your customers and complete insurance workflows end to end.
30 minute quotes now take 3.
Salman retweeted
In 1990, a team left Apple with ideas the world was not ready for.
They did not win commercially, but their thesis held true.
We’re doing the same for insurance, using language to cut the friction between intent and outcome.
Today, OpenSesame is becoming General Magic
@generalmagic_ai
Salman retweeted
After 5 years of coding with AI, I've written a comprehensive guide.
The only limit should be your own ideas: blog.silennai.com/claude-cod…
Salman retweeted
Shame @breadli428 got cut off, his work was super cool too: arxiv.org/pdf/2501.10100
Crazy that it transfers sim2real zero-shot.
@breadli428 doing a great job finishing his presentation after everyone got kicked out of the room at the Embodied World Models Workshop @NeurIPSConf three minutes into his talk...
#NeurIPS2025
Salman retweeted
One of the most engaging projects I’ve ever worked on. Excited to be attending NeurIPS this week with @_sshahid_ !
Our work on the @StanfordSVL BEHAVIOR Challenge revealed something unexpected: less reactive control modes performed significantly better.
Committing to 50-step plans (vs replanning every timestep) boosted success from 30% → 100% on subtasks like navigating & grasping.
Salman retweeted
Today, we present a step-change in robotic AI @sundayrobotics.
Introducing ACT-1: A frontier robot foundation model trained on zero robot data.
- Ultra long-horizon tasks
- Zero-shot generalization
- Advanced dexterity
🧵->