Founding Engineer: Manipulation

Robotics is the new paradigm. With hardware converging, Intelligence is the biggest bottleneck. Our world was built for humans and we're giving robots the ability to reason and act like us.

The way we learn any task - by seeing, doing, and feeling - is how we train robots. We're building general intelligence that works on any robot and fits seamlessly within our world.

We're building foundation models for deploying general-purpose manipulation robots.

Our data infrastructure generates rich, multi-modal human demonstration data at exponential scale. We use it to train manipulation models that work reliably across tasks, objects, and robot embodiments. We are live in deployment across multiple verticals.

The robotics industry is focused on building bigger, more generalizable pre-trained models, but a robot has no commercial value unless it works 99% of the time. Our research is focused on bridging the gap between 40% to 99% in real world deployments.

As founding engineer, you'll own the model pipeline end-to-end. You'll have $750K in compute credits across Azure, AWS, and GCP, NVIDIA Inception perks, and a bespoke data pipeline already running.

What you'll do:

• Own the train-test-deploy loop: curation, training runs, evaluation, deployment
• Select and tune architectures for our foundation model
• Collaborate with research organizations
• Set technical direction for the model team

You should have:

• Experience implementing models end-to-end on robots, or research experience in VLAs, world models, or imitation learning
• Proof of decision-making with incomplete information
• Extremely competitive nature
• No aversion to unglamorous work, in robotics, research and debugging have to go hand-in-hand
• A track record that shows you genuinely want to see robots deployed in the real world, projects, research, or competitions

Our culture is built on honesty and a willingness to fail fast. If you see yourself fitting here, please apply.

Comp:

• $80K-$200K
• 0.1%-2% equity
• Incredible office in NYC. Equipment budget, meals, etc.

We are developing foundational models for general purpose manipulation, trained on human demonstration data collected from our data factories – facilities where human data generators wear our data collection gloves and perform various tasks.

We use this data to train custom models that generalize over tasks, objects and work on any robot hardware.