Tanmay Agarwal
@tanmayxagarwal
Followers
412
Following
709
Media
9
Statuses
187
Helping build reliable robots @ Parallax Worlds | Prev Grad School @ Stanford in @StanfordSVL 🌲, Self Driving @NVIDIA, drones @senteraofficial
San Francisco, CA
Joined August 2019
After a year of building, I'm excited to introduce the world to Parallax Worlds. While robots bring immense ROI for automating hazardous and repetitive tasks, their reliability remains a critical challenge. Factory robots fail every 87 minutes on average, costing $1000s/min.
1
7
32
Getting interviewed by @Scobleizer was not on my bingo card this year! Proud of all the work that the team has been doing at Parallax. We’re hiring a Robotics Researcher and Backend Engineer. DMs are open!
The future of simulated worlds with founder of https://t.co/VsxIMEnd0L Teaches robots how to work on factory floors. At Llama Lounge, @jowyang’s San Francisco AI event.
0
1
11
Excited to see the research frontier build on Gaussian Splatting + Physics Engines as well! We've been big believers in this approach at Parallax, our website has a similar demo right on the landing page ( https://t.co/WF5yirRzP4) If you want to build enterprise grade simulations
parallaxworlds.com
Parallax is a modern reliability testing platform for robots. We build hyper-realistic robotics simulations to find robot failures before they occur in the real world.
Ever want to enjoy all the privileged information in sim while seamlessly transferring to the real world? How can we correct policy mistakes after deployment? 👉Introducing GSWorld, a real2sim2real photo-realistic simulator with interaction physics with fully open-sourced code.
0
2
6
We've had an overwhelming response to our Robotics Happy Hour for @Techweek_ today at 4 PM, with almost 400 requests to attend. We're bringing together a curated mix of top startups and investors shaping the future of robotics and AI. Our guest list includes engineers and
partiful.com
Join a curated group of robotics founders, engineers and investors in South Park. We'll start with a panel to discuss the current landscape of opportunities in robotics and have time for open...
1
5
12
If you have a video model that can truly generate realistic physics, you can solve autonomy for the trillion dollar manufacturing economy. Can we please try that out before we make social media apps again? Or give us access and we'll do it!
0
0
1
After 9 months of putting this together, I'm excited to finally share a project that I've wanted to exist for years now: The State of Adventure Capital
52
61
403
So the “Scale AI for robotics” is just Scale AI? What happens now 😂
2
1
12
Can attest, incredible work.
We just did World’s first on-stage autonomous demo of long-horizon dexterous VLA 🚨 No training. No setup. Performance out of the box. Live demo is hard and unpredictable, but we felt great about our model’s generalization, and it went pretty well! 💯 Zero-shot. 100% success.
0
0
7
(2/2) Side by side before & after of IsaacSim alone and ParallaxSim. This is really important because: - We can now eval perception-based algorithms in sim - We can eval robots in exactly their factory without having to build CAD models of the whole factory
2
2
9
One of my favorite examples of customers using Parallax is @BotBuilt_ & @cbames. BotBuilt is making robots to build homes and we brought their factory in NC into simulation with just a 20 minute video. (1/2) Parallax Sim that uses Gaussian Splats + Physics Engine (IsaacSim here)
1
14
123
Are you going to be at @foxglove Actuate next week? If so, come say hi, Parallax Worlds is excited to sponsor! We convert any space into a fully physics-enabled simulation environments for your robots to test & train in. We're also hiring across robotics, ML, & infra roles!
1
1
3
Given how much manual effort it takes, it's not surprising that most companies don't use sim at all or have 1-2 environments in which they test. At Parallax, we're building a product where we answer yes to every single question in the thread! Reach out if this sounds interesting
0
0
1
Congratulations, after all this you have set up one sim environment! Two issues to scale: - Defining and creating test cases is very manual -- Can we use agents+diffusion? - Need infrastructure to run them, aggregate metrics & logs -- can we build cloud infrastructure for this?
1
0
1
I now need to add my robot. Which brings us to 3: estimate friction, damping, stiffness for every joint of the robot and end effectors. People usually guess these values or use whatever "looks correct". Can we use System ID or newer methods to automatically estimate them?
1
0
1
1: We need CAD models for literally every single thing in the sim. Can we replace the background and every single object with Gaussian Splats? 2: Where do the objects go? Need the pose, mass, inertia, friction of all these objects. Can we estimate with agents?
1
0
1
One domain within robotics that I think is underrated with respect to how much modern AI can impact it is simulation. If agents / reasoning models are helping replace manual labor, there is so much that goes into running a sim that goes unnoticed: (thread)
1
1
4
🎉 Excited to share that our review paper on learning-based dynamics models for robotic manipulation is finally out in Science Robotics @SciRobotics! 🤖 Led by my former mentee @BoAi0110, this paper is especially meaningful to me. It builds on the structure of my PhD thesis and
7
38
252
the 10x activities can look like setbacks in hustle culture: - time with family - investing in real relationships - self care (exercise, hobbies) these are the highest ROI compounding moves.
1
1
8