Founding Robotics Simulations Engineer

Founding Robotics Simulations Engineer

Location: San Francisco | Full-time

 

ABOUT DRIFT

 

Drift (godrift.ai) is building the simulation infrastructure that helps robotics teams get robots from the lab into the real world 10x faster, reliably.

 

We work closely with teams building humanoids, mobile manipulators and industrial robots, on problems where physics, learning and real hardware meet.

 

We're a lean but fast paced, founder-led team. You'll own projects end to end, from the first conversation with a customer to the final robot deployment.

 

WHAT YOU'LL DO

 

• Build physics-accurate digital twins of robots, worlds and assets in MuJoCo, mjlab, Isaac Sim / Isaac Lab and whatever else fits the problem
• Take policies from sim to real hardware and back. Find the gaps, measure them and close them
• Design simulation scenarios with domain randomization, and turn the results into clear findings for customers
• Calibrate sim against real data with system ID and parameter estimation. A sim isn't done until it meets acceptance criteria on real measurements
• Work directly with customer engineering teams to scope ambiguous, research-heavy problems and ship real milestones on a timeline
• Help turn project work into reusable tools and infrastructure across the Drift platform

 

 

WHAT WE'RE LOOKING FOR

 

Core:

• Strong physics simulation fundamentals: rigid-body dynamics, contact and friction models, integrators, solver behaviour and timestep stability
• Hands-on experience with MuJoCo, mjlab, Isaac Sim / Isaac Lab, or another research or commercial simulator
• Robot modelling experience (URDF / MJCF / USD, kinematics, actuators, inertials)
• Sim2real or real2sim experience, ideally including a real deployment. Domain randomization done properly
• Comfortable working directly with hardware: you've brought up a robot, debugged a bridge, or chased down a sensor that lied

 

Nice to have:

• Deformable modelling: FEM, PBD/XPBD, MPM, cloth, rods and cables. You can reason about the trade-offs between accuracy, speed, stability and differentiability
• Real2sim tooling: Gaussian splatting, mesh cleanup, convex decomposition for sim speed, system ID, data-driven parameter estimation
• Reinforcement learning experience
• GPU-parallel simulation with Warp or MJX

 

Supporting skills:

• ROS2, OpenSim, Unity, Unreal
• Sensor models: camera and depth noise, IMU
• Linux, Docker, cloud computing

 

 

HOW YOU THINK

 

• You know why simulation is useful and where it lies. A sim can be trusted only when it has been validated against real data with clear acceptance criteria, and you build that validation in from day one
• You use the best agentic and AI tools to solve problems faster. You treat them as leverage, not a crutch, and you know when to check their output
• Extreme customer empathy. You want to understand the customer's robot, their deadlines and what a failure costs them, and you design around that
• You own outcomes, not tasks. High agency and proactiveness.
• You learn fast and communicate clearly with teammates and with customers' engineers
• You ship. Research-flavoured projects still get steady, visible progress every week, and something usable lands in a reasonable timeframe

 

 

WHY JOIN

 

• Work on real robots from leading robotics teams, not just toy problems or flashy demos
• Early-team ownership: your work shapes the product, the platform and the company
• You'll work at the point where physics, learning and real hardware meet, on some of the hardest open problems in robotics

 

 

HOW TO APPLY

 

Apply here or message us directly at founders@godrift.ai.

Along with your resume, send one short example of a sim you built or a sim2real gap you closed: what was off, how you measured it and what you changed. Links to code, videos or write-ups are welcome.