---
title: 'ML Infra Engineer, Modeling at Physical Intelligence'
canonical: 'https://feeny.ai/job/ml-infra-engineer-modeling-physical-intelligence-san-francisco-dt7e182daxxa'
type: 'job'
last_seen: '2026-09-06'
---

# ML Infra Engineer, Modeling at Physical Intelligence

- **Company:** Physical Intelligence
- **Location:** San Francisco, CA
- **Employment:** full-time
- **Work type:** onsite
- **Posted:** 2026-09-01
- **Last confirmed live:** 2026-09-06
- **Apply:** https://jobs.ashbyhq.com/physicalintelligence/96bf6ea8-fb8d-4a76-bec7-878a8c224cea

## Job description

Physical Intelligence is bringing general-purpose AI into the physical world. We are a group of engineers, scientists, roboticists, and company builders developing foundation models and learning algorithms to power the robots of today and the physically-actuated devices of the future.

In this role you will help scale and optimize our training systems and core model code. You’ll own critical infrastructure for large-scale training, from managing GPU/TPU compute and job orchestration to building reusable and efficient JAX training pipelines. You’ll work closely with researchers and model engineers to translate ideas into experiments—and those experiments into production training runs.

This is a hands-on, high-leverage role at the intersection of ML, software engineering, and scalable infrastructure.

The Team

The ML Infrastructure team supports and accelerates PI’s core modeling efforts by building the systems that make large-scale training reliable, reproducible, and fast. The team works closely with research, data, and platform engineers to ensure models can scale from prototype to production-grade training runs.

In This Role You Will

- Own training/inference infrastructure: Design, implement, and maintain systems for large-scale model training, including scheduling, job management, checkpointing, and metrics/logging.
- Scale distributed training: Work with researchers to scale JAX-based training across TPU and GPU clusters with minimal friction.
- Optimize performance: Profile and improve memory usage, device utilization, throughput, and distributed synchronization.
- Enable rapid iteration: Build abstractions for launching, monitoring, debugging, and reproducing experiments.
- Partner with researchers: Translate research needs into infra capabilities and guide best practices for training at scale.
- Contribute to core training code: Evolve JAX model and training code to support new architectures, modalities, and evaluation metrics.

## What We Hope You’ll Bring

- Strong software engineering fundamentals and experience building ML training infrastructure or internal platforms.
- Hands-on large-scale training experience in JAX (preferred), PyTorch.
- Familiarity with distributed training, multi-host setups, data loaders, and evaluation pipelines.
- Experience managing training workloads on cloud platforms (e.g., SLURM, Kubernetes, GCP TPU/GKE, AWS).
- Ability to debug and optimize performance bottlenecks across the training stack.
- Strong cross-functional communication and ownership mindset.

Bonus Points If You Have

- Deep ML systems background (e.g., training compilers, runtime optimization, custom kernels).
- Experience operating close to hardware (GPU/TPU performance tuning).
- Background in robotics, multimodal models, or large-scale foundation models.
- Experience designing abstractions that balance researcher flexibility with system reliability.

Pursuant to the San Francisco Fair Chance Ordinance, we will consider for employment qualified applicants with arrest and conviction records.

## About Physical Intelligence

## Company Overview
- **One-liner**: Physical Intelligence is building general-purpose AI foundation models that can control any robot to perform any physical task, aiming to bring the flexibility of large language models into the physical world.
- **Entity Type**: Private (Series A/B stage; total funding $1.07B across multiple rounds)
- **Headquarters**: San Francisco, California, United States
- **Founded**: 2024
- **Founders**: Sergey Levine (UC Berkeley), Chelsea Finn (Stanford), Karol Hausman (CEO, ex-Google DeepMind), Lachy Groom (ex-Stripe), Quan Vuong (ex-Google DeepMind), Brian Ichter, and Adnan Esmail

## Core Business
- **Primary industry**: Robotics foundation models / Artificial Intelligence / Research Services
- **Target customers**: B2B enterprise – logistics, grocery, manufacturing, and other verticals requiring physical automation (currently testing with a small number of partners)
- **Mission**: “Bringing general-purpose AI into the physical world” – developing learning algorithms and foundation models to power robots and other physically-actuated devices

## Products & Services
- **π0 (pi-zero)**: First generalist policy released in October 2024; a vision-language-action (VLA) model capable of dexterous multi-task robot control. Open-sourced in February 2025.
- **π0.5**: VLA with open-world generalization (April 2025); can control a mobile manipulator to clean unfamiliar kitchens or bedrooms.
- **π0.7**: Steerable robotic foundation model with emergent capabilities (April 2026); can control a new robot platform without additional training.
- **π*0.6**: VLA trained with reinforcement learning to improve success rate and throughput on real-world tasks (November 2025).
- **FAST**: Efficient robot action tokenizer enabling 5x faster training of generalist policies (January 2025).
- **Multi-Scale Embodied Memory (MEM)**: Gives models long- and short-term memory for tasks exceeding ten minutes (March 2026).
- All offerings are research-stage models; the company does not currently sell a commercial product but partners with select companies for testing.

## Market Standing
- **Valuation**: $5.6 billion (as of January 2026, per TechCrunch)
- **Total Funding**: $1.07 billion across four rounds:
  - Seed (March 2024): $70M led by Lux Capital and Thrive Capital
  - Series A (November 2024): $400M led by Lux Capital and Thrive Capital
  - Venture Round (November 2025): $600M from 8 investors
  - Venture Round (March 2026): undisclosed amount from 4 investors
- **Notable Investors/Partners**: Khosla Ventures, Lux Capital, Thrive Capital, Sequoia Capital, OpenAI, CapitalG, Redpoint Ventures, Bond
- **Growth Signals**: Headcount grew 183.5% year-over-year to 163 employees (LinkedIn, mid-2026); the company states it has “blown through” its original 5-10 year roadmap in just 18 months; test deployments in logistics, grocery, and food production.

## Competitive Advantages
- **Cross-embodiment learning**: Models can transfer knowledge to any new robot hardware without starting data collection from scratch, lowering the marginal cost of onboarding autonomy.
- **Pure research focus**: Unlike competitors (e.g., Skild AI), Physical Intelligence deliberately avoids near-term commercialization, allowing the team to pursue general intelligence without product pressure.
- **World-class founding team**: Combination of top robotics academics (Levine, Finn) and experienced entrepreneurs/operators (Groom, Hausman).
- **Open-source release**: π0 weights and code are publicly available, fostering community contributions and accelerating research.

## Strategic Focus
- Current priorities include scaling foundation models to more tasks and environments, improving generalization and robustness, and expanding partnerships for real-world testing.
- The company explicitly does not give investors a timeline for monetization; instead it focuses on building general-purpose physical intelligence with a 5-10 year horizon (though progress has been faster than expected).
- Future strategy centers on continuous improvement via a loop of data collection → training → evaluation → more data, rather than rapid deployment.

## Why Work Here
- **Culture**: Described as a “pure company” – internally driven by research needs, not external market demands. Highly collaborative, with a mix of engineers, scientists, and roboticists working together.
- **Work Environment**: Hybrid/in-office – most roles require on-site presence in San Francisco (396 Treat Ave), though some remote flexibility exists. In-office setting with a “no reception” vibe, open lab space with robot stations.
- **Growth**: Team has scaled from ~80 (Jan 2026) to 163 (mid-2026) and is still hiring across research, ML infra, hardware, and engineering roles. Plans to grow “as slowly as possible” to maintain quality.
- **Perks & Engineering Highlights**: Access to cutting-edge robotics hardware and compute resources; opportunity to publish research and open-source code; exposure to a wide variety of real-world automation challenges (e.g., robots learning to fold pants, peel vegetables, make espresso). Employees come from top institutions like Berkeley, Stanford, Google, NVIDIA, and Anduril.

## Sources
1. [physicalintelligence.company](https://www.physicalintelligence.company/) – Official website, model releases, and values
2. [TechCrunch](https://techcrunch.com/2026/01/30/physical-intelligence-stripe-veteran-lachy-grooms-latest-bet-is-building-silicon-valleys-buzziest-robot-brains/) – In-depth profile including valuation, strategy, and culture
3. [LinkedIn](https://www.linkedin.com/company/physical-intelligence) – Company details, headcount, funding, and talent sources
4. [Built In](https://builtin.com/company/physical-intelligence) – Career page, employee count, and office policy
5. [jobs.ashbyhq.com/physicalintelligence](https://jobs.ashbyhq.com/physicalintelligence) – Current job openings

## Other roles at Physical Intelligence

- [ML Infra Engineer, Data Systems](https://feeny.ai/job/ml-infra-engineer-data-systems-physical-intelligence-san-francisco-8g386zwjvd64) — San Francisco, CA
- [Software Engineer, Data Quality](https://feeny.ai/job/software-engineer-data-quality-physical-intelligence-san-francisco-p7f1n4g3j35e) — San Francisco, CA
- [Shift Lead](https://feeny.ai/job/shift-lead-physical-intelligence-san-francisco-23rb2r5nr82f) — San Francisco, CA
- [Fullstack Software Engineer](https://feeny.ai/job/fullstack-software-engineer-physical-intelligence-san-francisco-4d81qeq8e7hy) — San Francisco, CA
- [People Ops](https://feeny.ai/job/people-ops-physical-intelligence-san-francisco-k10k2h0mgp8j) — San Francisco, CA
- [NPI Technical Program Manager](https://feeny.ai/job/npi-technical-program-manager-physical-intelligence-san-francisco-8grzx6hq5pm4) — San Francisco, CA
- [Robot Operator](https://feeny.ai/job/robot-operator-physical-intelligence-san-francisco-gq9jsnxd3s5g) — San Francisco, CA
- [Manufacturing Engineer](https://feeny.ai/job/manufacturing-engineer-physical-intelligence-fremont-30v59z9v8ze2) — Fremont, CA
- [Production Test Engineer](https://feeny.ai/job/production-test-engineer-physical-intelligence-san-francisco-8ngeahchx0hx) — San Francisco, CA
- [Supply Chain Lead](https://feeny.ai/job/supply-chain-lead-physical-intelligence-san-francisco-9n0bq8qkcvzm) — San Francisco, CA