Member of Technical Staff, Robotics

Reactor; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Benefits offered by this job

Visa sponsorship

Relocation support

Health, dental, and vision coverage

Equity

Job summary

Reactor in San Francisco is building an inference platform for real-time world models. This hands-on role sets the technical direction of Reactor's robotics work from the ground up, ensuring customer models run reliably on real robots within strict latency budgets.

We're looking for a generalist across ML, systems and robotics who understands policy behavior at inference time, and how batching, scheduling and network paths contribute to end-to-end latency.

Qualifications

• Experience running learned policies or ML models on real robots or autonomous systems.
• Strong ML understanding of robot policies and world models at inference time, not training them.
• Depth on inference systems: batching, scheduling, GPU utilization, p99 latency vs throughput.
• Strong production-level C++ and Python coding skills.
• Experience with ROS/ROS2 and DDS in robotic middleware.

Responsibilities

• Serve robot policies on GPUs under robot-grade latency budgets and profile the full path from input to output.
• Build the integration layer between Reactor and robot stacks: ROS/ROS2 nodes, bridges, transport, and low-latency channels.
• Use world models for policy evaluation and tune rollouts for throughput and turnaround time.
• Bring teleoperation and human-in-the-loop experiences to real-time transport, including synchronization and failover.
• Debug with customer robotics teams across sensing, network, inference, and control loops.
• Help shape the technical roadmap and set engineering standards for the team.

Skills

C++

Python

ROS/ROS2

Latency budgeting

Robot policies

World models

Inference systems

GPU utilization

Tools

DDS

TensorRT

CUDA

Triton

Job description

Reactor is building the inference platform for real-time world models. Our goal is to democratize access to real-time intelligence to unlock next-generation use cases in media, physical AI and omni-directional models. We already run the largest and fastest catalog of world models on the market, and work closely with all major model labs to run and optimize their world models, including Alibaba, Amazon, and NVIDIA. Our team is made of alumni from leading firms such as Apple, Meta, Amazon, Microsoft, Luma AI, Adobe, Stability AI, and more. We've raised $74M led by Lightspeed, with Amplify Partners, WNDR, NVIDIA, Sapphire, and other Tier 1 investors participating. Jeffrey Katzenberg, co-founder of DreamWorks Animation, sits on our board of directors.

The Role

This is a hands-on role with a lot of scope. You'll help set the technical direction of Reactor's robotics work from the ground up. Day to day, you'll get customer models running reliably on our platform under strict latency budgets, wired into real robots.

We're looking for a generalist across ML, systems and robotics. You understand how a policy behaves at inference time, how batching, scheduling and the network path add to its end-to-end latency, and how its output tensor becomes an action in a robot's control loop.

What You'll Do

• Serve robot policies (VLAs, IL/RL models) on GPUs under robot-grade latency budgets, and profile the full path from input tensor to output action
• Build the integration layer between Reactor and robot stacks: ROS/ROS2 nodes and bridges, transport, and low-latency channels carrying observations up and actions down
• Use world models for policy evaluation, and tune how we serve rollouts for throughput, fidelity and turnaround time
• Bring teleoperation and human-in-the-loop experience to our real-time transport: synchronization, jitter and failover for a live control loop
• Debug with customer robotics teams across the whole loop: sensor capture, network, inference, action output, robot execution
• Help shape the technical roadmap and set the engineering bar for the team around you

What We're Looking For

• You've run learned policies or ML models on real robots or autonomous systems, and debugged failures from perception through policy, control and hardware
• Strong ML understanding. You know how modern robot policies (VLAs, imitation learning, RL) and world models are built (transformer backbones, action tokenization or diffusion heads, observation encoders) and how they behave at inference time. You won't be training them.
• Depth on inference systems: batching and scheduling, GPU utilization, p99 latency vs. throughput, and running models in real time against physical systems
• Strong C++ and Python in production, and comfortable changing performance-critical code
• Working experience with ROS/ROS2, DDS and robot middleware, from message definitions to transport QoS settings
• Range across ML, systems, cloud infra, networking and robotics. You'd rather fix whatever's blocking the robot than stay in one layer.
• The judgment to lead: setting direction, making tradeoffs explicit and representing Reactor in front of customer engineering teams

Bonus Points

• World models, video or diffusion models, or learned simulators for policy evaluation or training
• Teleoperation: low-latency control loops, sensor fusion for human-in-the-loop tools, demonstration data collection
• Inference optimization: INT8/FP8 quantization, CUDA/Triton kernels, TensorRT, vLLM/SGLang-style serving
• Real-time networking: WebRTC, RTP/SRTP, QUIC, jitter buffers, frame budgets
• Infrastructure: multi-tenant GPU serving, autoscaling, data pipelines, eval tooling
• Humanoid or dexterous manipulation
• Early-stage startup or founding-engineer experience

Logistics

We're in person in San Francisco

• Competitive SF salary and meaningful early equity
• Visa sponsorship and relocation support
• Generous health, dental, and vision coverage