ML Infra Engineer, Platform

The role

You will own and scale AI-native infrastructure, including Kubernetes clusters and microservice platforms. You will drive security, observability, and cost-aware infrastructure while hardening platform foundations and improving developer experience.

What you will do

• Operate and evolve Kubernetes clusters and service deployment patterns
• Build a scalable microservice platform for internal systems
• Support safe rollouts, upgrades, and rollback strategies
• Treat security, logging, metrics, tracing, and alerting as first-class platform primitives
• Improve cost visibility and enable cost-aware decision-making
• Design authentication/authorization flows, networking architecture, quota and rate-limiting services
• Reduce infra churn by standardizing patterns, abstractions, and interfaces
• Build clear, documented interfaces for using platform infrastructure
• Support consistent local vs. remote development workflows
• Work closely with researchers and other engineers to translate needs into reusable infrastructure

Required qualifications

• Extremely strong first-principles thinking
• Familiarity with agent infrastructure, observability, security and sandboxing
• Deep experience with cloud platforms (GCP, AWS) and distributed systems: compute orchestration, networking, autoscaling, service meshes, load balancing
• Ability to reason about system bottlenecks, performance tuning, and cost optimizations across compute, networking, and storage
• Comfort with Kubernetes, cluster-level reliability, and service-oriented architectures
• Solid intuition around scalability, performance, and failure modes
• Experience with infrastructure-as-code (e.g. Terraform), containerization, and modern platform engineering practices
• Familiarity with logging, metrics, tracing, incident response, SLOs, and debugging complex distributed systems
• Strong cross-functional communication and ownership mindset
• 4-6 years experience working in fast-moving or early-stage environments where ambiguity is normal

Nice to have

• Experience with large-scale ML training, evaluation, or simulation infrastructure
• Experience with secrets management systems (e.g., Doppler)
• Background in observability, cost optimization, or internal platform tooling
• Exposure to robotics, simulation, or real-time systems

About the company

Physical Intelligence is developing foundation models and learning algorithms to power robots and physically-actuated devices. The team builds the infrastructure for training VLA models, orchestrating simulation, and deploying intelligence across robot fleets.