Jobs at Genesis

Build scalable infrastructure for robotics simulation, including workload management, environment frameworks, task interfaces, evaluation tooling, and real-to-sim pipelines. Collaborate with policy training and asset-generation teams to improve sim-to-real performance. The role requires extensive physics simulation expertise, particularly rigid body simulation or FEM, along with CUDA and GPU programming proficiency. Experience with robotics simulation platforms, synthetic data, asset pipelines, manipulation, humanoids, or embodied AI is advantageous.

Operate robots through teleoperation equipment during onsite data collection sessions. Follow predefined tasks and standard operating procedures, monitor robot behavior, identify unsafe or unexpected activity, and maintain safety and consistency during repetitive sessions. Provide session feedback, remain focused for extended periods, and collaborate with operations staff and session leads. The role is hands-on and execution-focused, with technical setup and hardware maintenance handled by other teams.

Build and optimize distributed foundation-model training infrastructure across multi-node GPU clusters. Profile data pipelines, GPU kernels, memory, networking, and CPU/GPU workloads to improve convergence speed and utilization. Develop CUDA, cuDNN, Triton, and custom kernel solutions integrated with PyTorch. Create monitoring and debugging tools for large-scale training runs and diagnose performance issues.

Build and maintain the software engineering backbone of Genesis-World, an open-source robotics simulation platform. Responsibilities include scalable testing for JIT-compiled engines, deterministic reproduction and serialization, autodiff infrastructure, scene editing, extensible plugins, robot actuation, event systems, timing architecture, refactoring, performance optimization, and cross-platform debugging. The role also involves code review, issue triage, contributor support, and collaboration with compiler and physics teams.

Develop and optimize learning-based robotic manipulation and locomotion systems. Build low-latency teleoperation infrastructure, train reinforcement and imitation learning policies, deploy policies to real robots, diagnose control-pipeline bottlenecks, and reduce the simulation-to-real-world gap through simulation and hardware co-optimization.

Own Quadrants, an open-source Python compiler for performance-critical numerical computing. Responsibilities include API design, intermediate representation and lowering, multi-backend support, compiler optimization, GPU kernel development, numerical primitives, benchmarking, regression tracking, and roadmap ownership. The role combines GPU programming, compiler engineering, Python API design, performance optimization, user engagement, and open-source product direction, with potential to lead a small team.

Lead complex customer deployments of physical AI and robotics systems from prototype through production. Coordinate software, AI, hardware, infrastructure, operations, and customer teams; manage milestones, dependencies, site readiness, testing, validation, logistics, launches, risks, and technical blockers. Partner with enterprise customers, establish scalable deployment and release processes, and define approaches for scaling systems across diverse environments.

Design and optimize embedded control stacks for general-purpose robots, including motion planning, trajectory optimization, state estimation, and nonlinear optimal control. Build simulation-to-hardware software frameworks, integrate sensors and actuators, and debug and validate controllers on physical robots. The role requires production-grade C++ expertise, advanced robotics and control knowledge, and the ability to deliver reliable end-to-end robotic systems.

Designs and maintains foundational frameworks, internal tools, and reusable systems that accelerate AI research and engineering workflows. Collaborates across teams to validate use cases, simplify complex requirements, improve development processes, and reduce organizational friction. Responsibilities include identifying code design flaws, implementing large-scale changes, conducting code reviews, documenting practices, training teams, and maintaining high software quality.

Develop and ship production-ready physics simulation capabilities for Genesis-World, improving speed, fidelity, scalability, completeness, and solver coupling. Implement physics algorithms in Python using the Quadrants compiler across CPU and GPU backends. Validate methods through analytical solutions, cross-engine comparisons, and real-world data while supporting robotics and physical AI pipelines.

Own Quadrants, an open-source Python compiler for high-performance physics and numerical computing. Lead its API, IR, lowering, compiler optimizations, GPU and CPU backends, numerical primitives, benchmarks, and roadmap. Improve kernel performance, usability, correctness, and extensibility while partnering with Genesis users and the broader open-source community. The role combines GPU programming, compiler engineering, Python API design, performance analysis, and technical leadership, with potential to grow into managing a small team.

Design, build, and maintain petabyte-scale batch and streaming data pipelines for robotics foundation model training and evaluation. Own data models, storage systems, ingestion, transformation, and orchestration infrastructure. Standardize processing across teleoperation and synthetic simulation datasets while operating production-grade distributed systems and infrastructure.

Build and maintain the Genesis-World core engineering infrastructure to reduce friction for simulation: improve testing and JIT compile performance, enable bit-exact reproducibility, make autodiff efficient, support live scene editing and extensibility, improve actuation and time management, refactor core systems, and support community contributions and reviews.

Build a high-throughput GPU-based simulation pipeline for robotics foundation model training, implement actuators/sensors/controllers, bridge sim-to-real with policy training, and develop real-to-sim asset pipelines in collaboration with robotics and asset teams.

Lead profiling and performance ownership of a hybrid raster/ray-traced renderer. Profile, diagnose, and optimize GPU/CPU rendering across platforms, build custom instrumentation, evolve hybrid paths, set technical direction, mentor senior engineers, and collaborate with hardware vendors on low-level GPU issues.

Full-cycle recruiter responsible for sourcing and closing elite Research Scientists, Software Engineers, and Robotics engineers. Partner closely with founders and technical leaders, build scalable sourcing and screening workflows (Ashby/ATS), run a fast, candidate-friendly hiring process, and use data and automation to optimize pipeline health and conversions.

Lead development of a high-performance robotics simulation platform and compiler stack. Design compute infrastructure and dataflows for physics simulation and foundation-model training. Implement JIT/LLVM-based GPU codegen, optimize compile/runtime performance, add differentiable programming support, collaborate with simulation engineers, and contribute to open-source ML compiler projects.

Design and implement embedded control stacks, motion planning, and trajectory optimization for dynamic robots. Build real-time state-estimation pipelines fusing heterogeneous sensors, formulate and solve optimal control problems (nonlinear MPC, convex/trajectory optimization), create modular sim-to-hardware software, and lead on-robot debugging, tuning, and validation of controllers.

Develop and optimize a learning-based robotic manipulation control stack; design and maintain low-latency teleoperation systems; train and deploy policies using reinforcement and imitation learning; diagnose control-pipeline latency and sim-to-real gaps; co-optimize simulation and real-world robot behavior; collaborate on building general-purpose Physical AI.

Design, build, and optimize distributed PyTorch training systems for multi-node GPU clusters. Profile and eliminate performance bottlenecks across data pipelines, kernels, and networking. Implement low-level CUDA/cuDNN/Triton kernels, tune hardware-software interactions, and develop monitoring and debugging tools for large-scale training runs to maximize utilization and reduce time-to-convergence.