Physical Intelligence
Software Engineer, Robot Interfaces
San Francisco
Sponsorship not specifiedDetected 22 days ago
TypeScriptPythonReactFull-Stack DevelopmentAlgorithmsPostgreSQLGCPRESTgRPCWebSocketsRoboticsVoIPMedical DevicesResearch
Stay score
odds of building a lasting career here
37Risky
Cap-exempt (no lottery)0
Sponsors this role80
Entry-level history0
PERM / green-card track0
Lottery odds40
Fits your clock70
Thin sponsorship signal and lottery-bound. A low-probability bet with your clock running. Prioritize cap-exempt roles and proven entry-level sponsors first.
Lottery odds assume a STEM candidate.
Personalize to your clock →Community outcomes
No reports yet — be the first to help the next applicant.
About the role
- Physical Intelligence is bringing general-purpose AI into the physical world.
- You will work closely with operators, roboticists, researchers, and other engineers to turn physical workflows into reliable software.
Responsibilities
- Own the human-robot interfaces: The software people use to run, monitor, and debug robots, from setup and calibration through live operation and review. Someone lives in this for an entire shift, and small friction compounds fast.
- Design for the whole shift, not just the episode: Time on a robot is a fraction of the day.
- The goal is not to expose every tool, but to build one surface that tells someone what deserves their attention now and notices when they're stuck.
- Design for safety: Make state unambiguous, make failure obvious, and make the safe action the easy one. Assume the network drops mid-episode or a camera dies, and design so the system degrades predictably rather than surprisingly.
- Draw the line between device and cloud: Decide what has to stay local because it is safety-critical or latency-sensitive, and what belongs in a control plane that owns identity, state, and coordination across sites.
- Ship into the physical world: On-prem and on-device deployment, versioning and rollout across lab, warehouse, and partner sites, and enough observability to know what is running where.
- Act as your own PM: Gather requirements, prioritize, define success metrics, ship, and drive adoption. That starts by standing in the room and watching a shift.
- Strong fullstack engineering experience building production web applications and APIs, especially with React, TypeScript, and Python.
Requirements
- Experience with our specific stack: React, TypeScript, Python, Postgres, GCP, and Kubernetes.
Nice to have
- Strong UX instincts for high-stakes, high-frequency interfaces, where the user is standing up, under time pressure, and a mistake has physical consequences.
- Comfort designing workflows where people, hardware, and software have to function as one system.
- Strong product judgment, and experience working directly with users through ambiguous workflows and evolving requirements.
- Former founder, early employee, or other demonstration of comfort with ambiguity and solving hard problems end to end.
- Experience with teleoperation, robotics, or other robot-adjacent software.
- Experience with workflow, dispatch, or scheduling systems that direct human work in the physical world.
Skills
- The rest is preparation, troubleshooting, handoffs, and feedback, spread today across several tools and several people.
Benefits
- Experience with safety-critical or industrial systems (control room, medical device, aviation, automotive, or similar) where interface decisions carry real consequences.
Company info
- Fullstack builds the internal software the rest of Pi runs on.
- Our users span the path from research intent to physical execution: researchers designing experiments and requesting data, Research Ops working across task design, environments, prototyping, and instructions, and Production Ops running collection, evals, and deployments across our lab, warehouse, and real-world sites.
Apply directly at Physical Intelligence →Create a free account for alerts like thisView Physical Intelligence immigration profile
This listing is sourced directly from Physical Intelligence's careers page and normalized into a canonical job model.