About this Member of Technical Staff - Simulation Researcher role at Moonlake
Introducing Moonlake, AI for creating world simulations.
About Moonlake
Moonlake is building the frontier of interactive world models: systems that generate, simulate, and reason over 3D environments for robotics, embodied AI, and interactive applications.
We develop the infrastructure that enables intelligent systems to learn, evaluate, and interact within realistic virtual environments before operating in the physical world.
Our work sits at the intersection of:
Robotics
Embodied AI
Interactive 3D Worlds
World Models
Simulation Infrastructure
Physical AI
Moonlake is building the next generation of AI infrastructure for interactive digital worlds. Our mission is to enable anyone to create, simulate, and interact with rich environments using natural language and multimodal inputs, turning simple ideas into worlds with structure, physics, and intelligent behavior.
Our team has raised $28M in seed funding from NVIDIA Ventures, Threshold Ventures, AIX Ventures, and notable angels including Naval Ravikant and Jeff Dean to build the foundational layer for the future of AI—powering everything from robotics training and simulation to digital twins and interactive environments.
We are looking for exceptional researchers to advance the simulation technologies that will power the next generation of robotics and embodied intelligence.
The Role
We are looking for a simulation researcher to develop the physical models and numerical methods needed to simulate complex everyday phenomena that even the best current simulators struggle to reproduce. You will formulate these methods, implement them, and test their predictions against real-world measurements.
Examples include:
Dense simultaneous contact, such as many objects interacting during packing or manipulation.
Rigid–deformable-fluid coupling such as a zipper interacting with fabric, or water in a plastic bottle.
High-speed actuation under large forces and torques.
Cables (codimension two) and thin sheets (codimension one), including plastic bags, poly mailers, and resealable bags.
Plastic deformation, aging, and wear in paper and cardboard, as well as fracture and cutting.
Phase transitions in food, such as frying an egg, and the mixing of miscible fluids.
Wetting and absorption when liquids meet porous materials such as sponges.
Four priorities guide this research:
Efficiency. Robot training and evaluation require repeated trials. We value approximations that reduce computational cost while preserving the physical behavior that matters for the task.
Robotics co-development. You will work with roboticists to run comparable trajectories in simulation and on real robots. The goal is for the robot to sense the same forces and see the same motion in both.
Inverse physics and real-world calibration. You will use real-world data to determine which physical models and parameters reproduce observed behavior. The aim is a simulation that matches the data, rather than one that merely looks convincing.
Agentic simulation. You will investigate how an AI agent can construct a good physical simulation autonomously or with human guidance, from choosing and calibrating models to testing the code it writes.
What You'll Do
Develop simulation methods
Formulate and implement new models for contact, friction, deformable objects, fluids, phase changes, and coupled physical systems, drawing on the phenomena above.
Design discretizations, time integration schemes, and linear and nonlinear solvers that make these models stable, accurate, and computationally practical.
Creatively design simplifying assumptions that reconcile sim-to-real correlation and computational tractability.
Co-develop with robotics
Design experiments with roboticists that compare the same actions in simulation and in the real world, using trajectories, forces, contacts, and task outcomes as evidence.
Use those comparisons to identify which physical effects matter to the robot and to guide the next iteration of the model or numerical method.
Research inverse physics
Infer material properties, contact events, contact parameters, initial conditions, or other hidden quantities from video, sensor readings, and robot trajectories.
Account for noise and ambiguity in the measurements. Test whether the fitted parameters predict interactions beyond those used for calibration.
Automate simulation authoring with AI agents
Build agents that turn a task description and real-world observations into executable simulation code, selecting and composing the physical models and numerical methods developed above.
Research how to train or engineer these agents: how to help them choose appropriate assumptions and correct errors in the simulations they write?
Areas of Focus
Physical interactions
Dense contact and friction; multiphysics coupling; cables and thin sheets; plasticity, fracture, fluids, wetting, and phase change.
Numerical methods
Discretization and time integration; finite element method; numerical linear algebra; nonlinear optimization; stability, convergence, and efficient approximation.
Inverse and agentic simulation
System identification, calibration, uncertainty, validation against robot data, and methods that help agents construct and assess physical simulations.
What We're Looking For
A record of original physics simulation research, with publications at SIGGRAPH / ACM Transactions on Graphics, Journal of Computational Physics, or comparable venues.
Expertise in numerical methods for linear and nonlinear systems. You can reason about positive definite, semidefinite, and indefinite systems, select suitable solvers, and explain the tradeoffs.
A strong foundation in mechanics and physical modeling. You can derive governing equations, state the assumptions behind a model, and connect those assumptions to observable behavior.
Experience implementing simulation algorithms in C++ and/or Python and evaluating their accuracy, robustness, and performance through careful experiments.
Ability to work with roboticists to design experiments and use the results to test and improve physical models.
Interest in how AI agents can help construct and calibrate simulations.
Why This Role Matters
Many interactions that robots encounter every day are still difficult to simulate well. Progress requires models that predict what robots encounter and numerical methods fast enough for large-scale evaluation and training. Your research will expand the range of phenomena Moonlake can simulate and help turn those advances into simulations that roboticists, engineers, and AI agents can use with confidence.
We are committed to being an on-site, in-person team based in San Francisco.
We are committed to being an on-site, in-person team currently based in San Francisco.