Software Engineer, Robotics Simulation & AI Infrastructure

About Qualcomm Robotics

The Qualcomm Advanced Robotics Team is building the AI-first stack for the next generation of general-purpose robots — from AMRs and cobots to emerging humanoids — pairing heterogeneous compute (CPU/GPU/DSP/NPU) with a full Robotics SDK, an integrated simulation platform, and AI operations infrastructure. With our high-performance robotics SoCs, workloads that previously required the cloud now run on-device, at the edge, at scale.

About This Team

We build AI infrastructure for modern robotics. The simulator is a central tool in it: a modern platform developed in-house, built on high-performance computing and open data formats such as OpenUSD, URDF, and glTF, intended to be best-in-class, and tuned for Qualcomm robotics SoCs and the robots they run on. The robots span the field — tabletop manipulation with robotic arms, legged locomotion, navigation and vision, and long-horizon tasks that carry a robot through multi-stage scenes requiring reasoning over many steps. This is a ground-up software effort: runtime architecture, performance, clean APIs, and developer tooling.

The simulator is a first-class target on both developer workstations and cloud compute — interactive authoring and vectorized GPU environments locally, orchestrated headless jobs at scale in the cloud. We use it to train policies, generate synthetic data for robot foundation models, gate software releases in CI, run hardware-in-the-loop against real silicon, and close the sim-to-real gap on deployed robots. We deliver jointly with Qualcomm’s AI operations workstreams, and a meaningful slice of the engineering sits where simulation plugs into the training, dataset, and deployment infrastructure they own.

The Opportunity

You will own significant pieces of a simulation platform on the critical path of Qualcomm’s robotics products — the runtime and its abstractions, the physics and rendering integrations, the training and data-generation paths, the HIL and CI plumbing, or the developer experience that makes all of it usable by robotics engineers who are not simulation specialists — across robot arms, legged platforms, and mobile robots.

This is first and foremost a simulation role — roughly 80% of the work is the simulator itself as a critical platform component. The remaining ~20% is AI infrastructure enablement: integrating simulation into the training, dataset, and deployment pipelines owned by the AI operations team, working with them rather than replacing them.

The role is open from mid-level through senior/staff, roughly 2 to 10+ years. We calibrate level from your depth during the interview loop, so apply if you are anywhere in that range.

What You’ll Do

• Design and build core simulator subsystems — scene representation and authoring, physics backends, sensor models, rendering — across the scenarios our robots work in: tabletop manipulation, legged locomotion, navigation and vision, and long-horizon, multi-stage tasks.

• Profile and optimize simulation and training workloads — physics solvers, rendering, data pipelines — so interactive workstation sessions and training throughput at scale stay high.

• Integrate and extend best-in-class open engines — GPU physics, USD/Hydra rendering, offline path tracing — behind clean, swappable interfaces.

• Build high-throughput paths for policy training and synthetic data generation: vectorized environments, GPU-resident state, procedural scene variation, domain randomization, and ground-truth labeling — consistent from an interactive workstation session to scheduled headless runs in the cloud.

• Make simulation a release gate: scenario suites, deterministic replay, benchmarks, and metrics that catch regressions before they reach hardware.

• Enable the continuous learning workflows that simulation feeds into, in partnership with the AI operations team: make simulation a first-class, config-driven stage in collect–merge–train–eval pipelines with reproducible run manifests and artifacts, so the same experiment runs locally and as containerized GPU jobs (Argo Workflows on Kubernetes) without the sim engineer owning the surrounding infrastructure.

• Stand up hardware-in-the-loop configurations: production robotics software running on Qualcomm silicon in communication with the simulator on the host workstation or cloud, and quantify where simulation and reality diverge — system identification, contact and actuator modeling, sensor noise, measured transfer results.

• Partner with AI operations, perception, controls, and silicon teams, and own the design docs, reviews, tests, and APIs others depend on.

Who We Look For

• Strong software engineering fundamentals, with production code you can talk through in depth.

• A working grasp of simulation as applied to robotics: rigid-body dynamics, kinematics, coordinate frames, numerical integration, or sensor models. — or the aptitude to build one quickly; engineers from games, graphics, and ML infrastructure backgrounds ramp well here.

• You pick up unfamiliar stacks quickly and independently, and you use modern AI tooling effectively in your development loop.

• You would rather converge two half-parallel code paths than add a third, and you are comfortable where some layers are settled design-of-record and others are still being validated.

• You are motivated by physical AI: work that ends in a robot that functions, not only a benchmark number.

Depth in two or three of these matters more than familiarity with all of them. If you're excited about this role but don't match every qualification, we encourage you to apply.

• Robotics simulators or real-time 3D engines: MuJoCo/MJX, Isaac Sim or Isaac Lab, Newton, PhysX, Gazebo, Drake, Unreal Engine, or Unity. Candidates from games, VFX, or AV simulation are welcome — these skills transfer directly.

• GPU programming and performance engineering: CUDA, Warp, Vulkan, compute shaders, or accelerator-aware data layout.

• Machine learning for robotics: reinforcement learning at scale, imitation learning, robot foundation models and VLAs, PyTorch.

• 3D graphics and scene pipelines: OpenUSD composition, PBR materials, real-time and offline rendering, sensor simulation.

• Robotics and distributed systems: ROS 2/DDS, URDF/MJCF, pub/sub and schema-driven wire formats, containers, cluster orchestration, and CI/CD for compute-heavy workloads.

• Working fluency in the AI infrastructure around simulation — enough to integrate with it and partner effectively with the AI operations team (who own it), not to run it yourself: Kubernetes-native workflow orchestration (Argo Workflows or similar), containerized GPU jobs and shared-storage datalakes (NFS/PVC volumes), config-driven experiment frameworks, checkpoint/resume semantics, and experiment tracking (TensorBoard, Weights & Biases).

• Effective leverage of modern AI tools as a core engineering skill: frontier AI models, AI coding and agentic workflows, AI-assisted learning for ramping on new domains quickly, and building knowledge bases — personal or team — to accelerate on-boarding and future work.

Principal Duties and Responsibilities

• Designs, develops, integrates, and validates the in-house simulation platform and its tooling, targeting Qualcomm robotics SoCs and robots across form factors and scenarios — tabletop manipulation, legged locomotion, navigation and vision, and long-horizon multi-stage tasks.

• Specifies environments, tasks, and evaluation criteria for training and assessing robot policies; partners directly with robotics researchers and engineers to iterate on tooling, workflow, and model performance until results hold up on real hardware.

• Builds digital twins of robots, sensors, and work cells, and hardware-in-the-loop configurations where the robotics stack runs on Qualcomm silicon against a simulator on the host, so algorithms are developed and validated before hardware time is committed.

• Builds synthetic data pipelines for robot foundation models and task models, and simulation-based regression suites that gate software releases.

• Works with the in-house sensor teams to model sensor behavior from real characterization and calibration data, delivering high-fidelity sensor simulation modules.

• Works closely with the robotics AI infrastructure team to complete the data flywheel: a continuous learning pipeline spanning data generation, training, evaluation, deployment, and back to data generation.

• Performs code reviews and regression tests; triages and fixes issues; at senior levels, leads design and testing efforts.

• Collaborates with hardware, systems, test, and AI operations teams so simulation integrates with the rest of the program; at senior levels, engages external partners and industry collaborations to advance simulation technology and shape shared roadmaps.

• Writes and reviews technical documentation, design rationale, and knowledge-sharing material for complex software projects.

• At senior levels, articulates technical findings and engineering due diligence — measured capabilities, trade-offs, and honest limits — to leadership, informing roadmap and strategy decisions.