Software Engineer, Data Quality

Physical Intelligence; United States; On-site; USD 90,000 - 150,000; Full time

Job summary

Physical Intelligence is building general-purpose AI-enabled robotics software in San Francisco, bringing data-driven QA to training datasets. This role focuses on QA and audit of multimodal data from vendors to ensure quality before model training.

You will partner with researchers to define concrete, measurable quality rubrics, build the tooling to automate sampling and validation, and coach vendors to meet internal standards.

Qualifications

• ~5 years of experience across building customer facing software.
• Vendor management with third-party data providers.
• Strong ops acumen with data-driven decision making.
• Hands-on in data: reviewing samples, writing rubrics, and auditing outputs.
• Clear written and verbal communication with researchers, ops teams, and partners.

Responsibilities

• Own external data QA & audit: build end-to-end QA for data from vendors, with sampling, review workflows, pass/fail decisions, and audit trails.
• Define quality standards with researchers and translate them into measurable rubrics.
• Build software interfaces, tooling, and infrastructure to scale automated data QA and track data quality.
• Manage vendor quality loops: provide feedback to vendors and align QA standards between vendor and internal teams.

Skills

Vendor management

Ops analytics

Data QA

Hands-on data review

Communication skills

Software development

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 learning algorithms to create a model that will control any robot to do any task.

The Team

Training robot foundation models takes enormous amounts of multimodal data. The Data Operations team makes sure that data is high quality, well understood, and safe to train on. This role owns quality assurance and audit for data, working closely with the researchers who consume it and the teams who create it.

In This Role You Will

• Own external data QA & audit: Build and run the end-to-end QA process for multimodal data ingested from third-party vendors and partners — sampling strategy, review workflows, pass/fail decisions, and audit trails — so data is verified before it reaches training.

• Define quality standards: Partner with researchers to translate what "good data" means into concrete, measurable rubrics and acceptance criteria; keep them versioned and current as research needs evolve.

• Build the system: Build the software interfaces, tooling, and infrastructure scale the system in an automated way - correctly sampling the right data for QA, internal QA checks, tracking dips in data quality from vendors and exposing those insights back to the vendors.

• Manage vendor quality loops: Deliver structured, actionable feedback to vendors; coach them on standing up QA checks on their own side; iterate until vendor-side and internal QA agree; enforce standards through review and acceptance processes.

What you'll bring

• ~5 years of experience across building customer facing software.

• Vendor management experience: working directly with third-party data providers, delivering actionable feedback, and measurably improving the quality of their output.

• Strong ops acumen and drive — a bias to instrument, measure, and iterate quickly rather than crank hours at a broken process.

• Comfort being hands-on in the data: reviewing samples yourself, writing and refining rubrics, and spot-checking your own program's output.

• Clear written and verbal communication across researchers, operations teams, and external partners.

Nice to have

• Experience with robotics, egocentric video, or other multimodal sensor data.

• Experience defining requirements for QA/validation tooling or working with engineers on automated data quality checks.

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