Parav Nagarsheth

Now (Updated April 2026)

I'm working on building infrastructure that helps us get to billions of useful robots.

Hot (and some lukewarm) takes

→
Robotics won't be solved unless we also solve sim2real Teleoperation, egocentric, UMI or real world RL won't work for non-humanoid form factors and tasks that require interacting with heavy, fragile or hazardous objects.
→
Exoskeleton (UMI) and egocentric data will get us close Human-collected data will automate a lot of valuable tasks. Scaling diverse, high-quality data collection at low $ / hour will be key to early breakthroughs
→
Most robots at foundation labs are limited by poor hardware Industrial robot arms have no regard for reflected inertia, backdriving efficiency, and torque transparency. Off-the-shelf QDD actuators have poor performance characteristics.
→
We're near the physics ceiling of current actuator tech Motor and gearbox architecture scaling is largely saturated. The remaining levers are thermal management and materials science.
→
The Market will force multi-embodiment once component hardware matures Robots are essentially wrappers around actuators after all.
→
Startups will build custom, weird-looking, task-specific robots for each $1B vertical. Like this.

Previously

I founded Fondant to build AI-native tools for completely new interactive experiences. We tried a bunch of things that didn't work, but I'm extremely proud of:

Before that, I was building and consulting on ML engineering projects — early diagnosis of respiratory diseases, AI for design at Diagram (acquired by Figma), and post-training early LLMs with EleutherAI.

Earlier still, I was an AI research scientist at Pindrop, working on account takeover fraud prevention, voice deepfake detection, and speech synthesis. Publications here.

I graduated with EE/CS degrees from Georgia Tech and Nirma University. At Nirma, I was part of the team that won Robocon India.

Bets on the future

01 World-class education that finally scales
02 Self-driving cars 10× safer and 2× faster than human-driven ones
03 Drones — wheeled and flying — doing package delivery
04 Robots that can manufacture, build infrastructure, and cook you Michelin-starred meals
05 Near-zero solar energy prices making synthetic fuels viable
06 Fundamental breakthroughs in aging and longevity
07 Airships dominating transcontinental cargo

Say hello

I'm on Twitter/X — that's the best way to reach me.