Member of Technical Staff - Product Engineering
The Product Engineering team builds the platform that lets other companies use PI's models: access to our models and the services around them, so a robotics company can build on PI the way developers build on API-based LLMs. See The Physical Intelligence Layer.
Concretely, you own the product surface our partners touch: they send us data, get a model back, and run it. That surface is remote inference and the inference client, partner data ingestion and APIs, and deployment integrations. You take a new partner from raw data to a deployed, evaluated model with little hand-holding.
In This Role You Will
Build the platform that lets other companies use PI's models: give partners access to our models, fine-tuning, remote inference, and the services around them, so a robotics company can build on PI the way developers build on API-based LLMs. This spans data ingestion and APIs, a partner portal, and deployment integrations, all working end to end and self-serve.
Ingest partner data end to end: take a new data or robot partner from their first sample to featurized, validated data in our system, and to a checkpoint they can eval.
Deploy and serve partner models: stand up remote inference endpoints, validate them, and get partners running our policies at low latency in their own environment.
Be the engineer embedded in partner engagements: sit in the partner channel, debug their deployment across the full stack, unblock them, and translate what they need into what we build.
Write production-quality code that interfaces with PI's infrastructure.
Bridge research and partners: turn research advances into deployable systems, and surface real-world failure modes back to researchers and engineers.
What We Hope You'll Bring
This is a software and systems role first, so a robotics or ML research background is not required. We care more about what you can do than whether you fit a standard profile.
An exceptional generalist software engineer who ships fast and owns results end to end. You have strong backend and systems design instincts, you understand how to run inference with our models, and you do your best work directly with partners and researchers.
Strong engineering skills: clean Python, the ability to interface with infrastructure, and sharp debugging instincts.
Strong backend and systems design: you can design a scalable system (databases, caching, APIs, services) and defend it under scrutiny.
Enough ML to run inference: you understand how to deploy, serve, and debug our models, even if you do not train them.
A practical, ownership mindset: you are motivated by making things work end to end.
Clear communication with researchers, operators, and partners.
Comfort with ambiguity and with on-site, embedded partner work.
Bonus Points If You Have
Founded or worked at an early-stage robotics, AV, or infrastructure startup.
Low-latency and real-time networking experience (inference transport, streaming, QUIC or websockets).
Experience with robot manipulation platforms, VLAs, or other ML models.
Familiarity with our stack: Python, Postgres, ClickHouse, GCP, Kubernetes, Modal, React and TypeScript.