We got into @ycombinator.
We're a robotics company that's betting against robotics.
jk... but,
when we watched robots actually get deployed.
Every "autonomous" one had a person behind it like 1x, Zoox, Waymo, almost all you've heard.
So to bet our money on the opposite: the human is the product that takes robotics foundational models to the next level.
@aktoriarobotics is the remote-operations layer for robots teleop software + trained operators, when a robot get's stuck, one of ours takes over in under a second.
cuz someone has to run the 10% autonomy can't reach yet
We are waiving the onboarding fee for our next 10 customers :)
We got into @ycombinator.
We're a robotics company that's betting against robotics.
jk... but,
when we watched robots actually get deployed.
Every "autonomous" one had a person behind it like 1x, Zoox, Waymo, almost all you've heard.
So to bet our money on the opposite: the human is the product that takes robotics foundational models to the next level.
@aktoriarobotics is the remote-operations layer for robots teleop software + trained operators, when a robot get's stuck, one of ours takes over in under a second.
cuz someone has to run the 10% autonomy can't reach yet
We are waiving the onboarding fee for our next 10 customers :)
We're a @ycombinator-backed robotics company.
We spend our weeks talking to the people deciding what data the world's robots learn from.
If I were getting into robotics and physical AI, these are the people I'd follow:
FRONTIER LABS
Ben Levin: director, robotics & physical AI data @NVIDIARobotics
Svetlana Grant: AI partnerships, AI for robotics @GoogleDeepMind
Yansong Pang: head of robotics AI partnerships @AIatMeta@peterxichen: head of frontier AI & robotics, Amazon FAR (ex-Covariant CEO)
@Vivan_Amin: principal research TPM @MSFTResearch, shipped rho-alpha
@alexttoshev: research director, @wayve_ai Labs
ROBOT COMPANIES
Andrew Barry: co-founder & CTO @GeneralistAI
Connor Keane: data ops, Atlas @BostonDynamics
Arash Momeni: head of AI data collection, Humanoid
Jonas Messner: head of AI @NEURARobotics
Mouli Sivapurapu: head of data & applications eng @Mind_Robotics
Pally Kumar: head of ops, Genesis AI
Tue Vu: research ops, Field AI
Sandeep Mariserla: head of data ops, Odyssey
Notice the pattern? Half of this list runs data. The bottleneck in robotics isn't models anymore. It's real-world robot data at scale.
That's what we're building at Aktoria (YC S26).
Who else would you add?
We're @aktoriarobotics (@ycombinator S26). We make robots work before they're fully autonomous.
AI learned from the internet. Robots need real-world experience.
We provide:
→ Low-latency teleoperation
→ 24/7 trained robot operators
→ Real-world training data
Humans are the API for robots.
Building robots?
What's your biggest autonomy bottleneck?
The next billion-dollar robotics company might not build robots.
It might build the infrastructure that keeps them running.
At @aktoriarobotics (YC S26), we're building remote teleoperation and real-world training data for Physical AI.
Building robots, scaling fleets, or training foundation models? Let's talk.
We're a @ycombinator-backed robotics company.
We spend our weeks talking to the people deciding what data the world's robots learn from.
If I were getting into robotics and physical AI, these are the people I'd follow:
FRONTIER LABS
Ben Levin: director, robotics & physical AI data @NVIDIARobotics
Svetlana Grant: AI partnerships, AI for robotics @GoogleDeepMind
Yansong Pang: head of robotics AI partnerships @AIatMeta@peterxichen: head of frontier AI & robotics, Amazon FAR (ex-Covariant CEO)
@Vivan_Amin: principal research TPM @MSFTResearch, shipped rho-alpha
@alexttoshev: research director, @wayve_ai Labs
ROBOT COMPANIES
Andrew Barry: co-founder & CTO @GeneralistAI
Connor Keane: data ops, Atlas @BostonDynamics
Arash Momeni: head of AI data collection, Humanoid
Jonas Messner: head of AI @NEURARobotics
Mouli Sivapurapu: head of data & applications eng @Mind_Robotics
Pally Kumar: head of ops, Genesis AI
Tue Vu: research ops, Field AI
Sandeep Mariserla: head of data ops, Odyssey
Notice the pattern? Half of this list runs data. The bottleneck in robotics isn't models anymore. It's real-world robot data at scale.
That's what we're building at Aktoria (YC S26).
Who else would you add?
If you’re applying for an O1 Visa in Robotics
Drop your questions in this thread
Our founders both got their O1s approved this month
@karanwalia_ai and @ShreenabhA
We got accepted to @ycombinator in summer 2026 and are now working with US based robotics founders, Frontier and Research Labs.
@Accel@CarnegieMellon
We're @aktoriarobotics (@ycombinator S26). We make robots work before they're fully autonomous.
AI learned from the internet. Robots need real-world experience.
We provide:
→ Low-latency teleoperation
→ 24/7 trained robot operators
→ Real-world training data
Humans are the API for robots.
Building robots?
What's your biggest autonomy bottleneck?
Aktoria Daily Takeover 🥳
Has your robot frozen today? How about a human taking over in under a second?
Please keep your robots working, and keep a human on call. https://t.co/xxwRDVOHRr
Flora is a suite of open models: it parses the physical world AND turns raw footage into rich training signals.
Flora delivers Astra-level performance at 125x lower cost, and the weights are open 🔥
Congrats to @DrJimFan and the NVIDIA GEAR team
Open weights like these still need real robot data underneath. That's Ben Levin's world at @NVIDIARobotics (Director, Robotics & Physical AI Data).
For anyone claiming robotics data is a solved problem or already commoditized:
- Confuses egocentric data with all types of robot-training data
- Has never tested a robotics model zero-shot
- Has never experienced the pain (and joy) of overfitting
- Has never deployed a policy on real hardware, not just in sim
- Has never tested generalization beyond the tasks seen in training or standard benchmarks like robodojo
Robotics data is far from solved.
free robot data you can download and check yourself before you pay.. that's rare in physical ai
@ShreenabhA and team at @aktoriarobotics just opened R1 Teleop samples
what it is
> real-robot teleop episodes recorded on their own station
> egocentric data from human workers, stereo + IMU
> free samples, custom collection on request
how it works
when a deployed robot freezes on the unfamiliar part of a job, a trained operator takes over remotely
every takeover gets recorded
that becomes training data
claimed vs proven
> samples page lists 16 embodiments, joints + gripper force at 50 Hz, MCAP / ROS 2 format
> latency numbers (~130 ms india to SF) are their own tests, no independent benchmark yet
> the launch video is only human egocentric footage.. looms, garments, assembly. the robot episodes are on the samples page
founders: Shreenabh (ex-Zoox, CMU, IISc) and Jaskaran Singh Walia, YC S26. no round size published that i could find…
#MachinesThatThink
We built a teleop team that nothing can stop.
Bad Wi-Fi? Stuck grippers? If the bot can move, our operators will move it, even if it’s an entirely new robot body.
Meet the Aktoria Robot Operations Center: https://t.co/xxwRDVPfGZ