ML Research Engineer - Robotics

D24 Search Limited; San Francisco (CA); On-site; USD 270,000 - 330,000; Full time

Benefits offered by this job

Equity

Job summary

D24 Search Limited is building the US presence in San Francisco to lead the data layer for physical AI across robotics and world models.

You’ll be the first robotics-focused researcher, training robot policies on proprietary synthetic data, measuring improvements, and turning results into both public demonstrations and product requirements. You report to the VP of R&D and collaborate with global teams.

Qualifications

• Trained and evaluated robot policies in simulation (Isaac Sim / MuJoCo or equivalent).
• 6+ years hands-on robot learning experience in industry or academia.
• Experience with policy learning methods: VLA fine-tuning, imitation learning, or RL.

Responsibilities

• Train policies on our data and measure improvements in manipulation tasks.
• Establish baselines on public eval suites and report deltas.
• Translate experimental findings into data specifications for the data team.
• Publish and demo results to the robotics community and participate in customer discussions.
• Hire and lead the Bay Area robotics team as it scales.

Skills

Robot learning

Experiment design

Policy learning

Simulation (Isaac Sim/MuJoCo)

ML tooling (PyTorch)

Education

MS or PhD in robotics or ML

Tools

Isaac Sim / MuJoCo

Job description

Founding ML Research Engineer - Robot Learning

Up to $300K + Equity

My client is building the data infrastructure layer for physical AI — the next frontier of artificial intelligence where systems understand three-dimensional space and the physics of the real world. From robotics to world models, the AI labs pushing the boundaries of what machines can perceive and interact with need physics-accurate synthetic training data, and that's exactly what we provide.

Today, teams at Meta Reality Labs, World Labs, and Epic Games already work with us and the company is an NVIDIA Inception member. We closed a ~$10M seed round led by Inovia Capital with Insight Partners and USVP, and is actively working toward a $100M+ Series A in early-to-mid 2027.

We are now building its US presence in San Francisco, where the vast majority of its customers and the physical AI ecosystem are based. The leadership team plans to relocate to SF after the Series A, making it a true headquarters — not a satellite office.

The company operates with a small, engineering-heavy team and a deeply entrepreneurial culture where autonomy, speed, and scrappiness are valued above all else.

What we are looking for.

We build the data layer for physical AI: we evaluate world models and robot policies against physics and generate the synthetic training data that fixes what breaks.

Today, the proof that our data makes robots better lives inside our customers' labs.

You'll bring that proof in-house as our first robot learning hire. You'll train robot policies on our data, measure what improves, and turn the answer into both a public result and a product requirement. You report to our VP of R&D and work daily with the engineering and product teams in Paris and the US team forming around you in the Bay.

You'll feel at home here if you'd rather design the experiment yourself than wait for someone to hand it to you.

What You'll Do

• Train policies on our data. VLA fine-tuning, imitation learning, or RL inside Isaac-class simulation, starting with manipulation and grasping.
• Measure the delta. Establish baselines on public eval suites, retrain with our data under the same seeds and protocol, and report what moved and what didn't.
• Turn results into specs. Write the requirements for the data team: what to generate, at what fidelity, with what variation.
• Publish and demo. Share results with the robotics community and join customer technical conversations.
• Build the team. Start as an IC and hire the Bay Area robotics team as it grows.

What will you be doing?

• Train robot policies (VLA fine-tuning, imitation learning, diffusion policy) on proprietary generated synthetic data and run rigorous experiments to quantify where the data moves the needle on manipulation and grasping tasks
• Establish baselines on public evaluation suites and open policy architectures, then measure deltas when our data is added — reporting results honestly, including failures
• Translate experimental findings into written data specifications and requirements for the data team in Paris, driving what to generate, at what fidelity, and with what variation
• Publish and demo results to the robotics community (CoRL, RSS, ICRA) and participate in customer technical conversations
• Hire and lead the Bay Area robotics team as it scales around you

Requirements:

• Trained and evaluated robot policies in simulation (Isaac Sim / Isaac Lab, MuJoCo, or equivalent)
• MS or PhD in robotics, ML, or related field
• Policy learning methods: VLA fine- tuning (OpenVLA, π0, GR00T), imitation learning (diffusion policy, ACT), or RL in sim
• 6+ years of experience in hands- on robot learning; industry and university research lab both count
• Used eval results to set the data or sim changes for the next training iteration
• Graduate degree from a leading robotics program (CMU Robotics called out as a green flag)
• Ran controlled comparisons: fixed seeds, checkpoints and eval suite, varying only the training data

Nice-to-have

• Manipulation or grasping policy training
• Navigation policy training in simulation
• True groundbreaking and rapidly expanding area for Physical AI
• Competitive base salary ($260 – $300K+ depending on experience)
• Meaningful early-stage equity (founding-level grants with significant upside ahead of a $100M+ Series A)
• San Francisco office space (near Caltrain for accessibility)
• Opportunity to be a founding US team member with VP-track trajectory
• Direct access to founders and leadership team