Overnight Gpu Programming Jobs (NOW HIRING)

About the Role

We build robots that run 24/7 in production environments. We're looking for a hands-on engineer to own the reliability, infrastructure, and developer tooling that keeps our fleet running and our engineering team fast. You'll split your time between robot-side systems work, cloud infrastructure, and building automation that multiplies the team's output.

A significant portion of this role involves working with AI coding agents. You'll direct autonomous agents to diagnose CI failures, triage production issues, run automated security and compliance checks, and execute multi-step engineering tasks. Knowing how to scope work for an agent, review its output critically, and build tooling that agents can use effectively is as important as writing the code yourself.

What You'll Do

• Own robot-side software (Python): Maintain the on-robot codebase that orchestrates arms, cameras, sensors, and I/O. Debug production hardware/software failures and ship fixes fast
• Build and maintain infrastructure as code: Manage cloud infrastructure — identity and access management, CI/CD credentials, secrets, container registries, cluster autoscaling — using declarative configuration and reproducible builds
• Drive build system and packaging migrations: Own the transition of robot software packaging to reproducible, hermetic build systems. Maintain machine images, dev environments, and deployment pipelines
• Build simulation and testing infrastructure: Develop end-to-end simulation systems that validate robot behavior without physical hardware — camera projection, kinematics, placement validation, fleet-wide calibration
• Develop and operate AI-powered engineering automation: Build autonomous agents that run nightly CI triage, security audits, infrastructure compliance checks, and code quality sweeps. Design the interfaces and instructions that make agents effective at real engineering work
• Improve observability and health monitoring: Instrument robot software with metrics and structured telemetry. Build alerting that catches problems before humans notice them
• Work across the stack: Touch frontend, backend, protobuf definitions, deployment tooling, and cloud services as needed. No part of the system is someone else's problem

What We're Looking For

• 3+ years of Python in a systems context — not web/ML Python, but the kind where you deal with processes, hardware I/O, async, and real-time constraints
• Strong Linux systems knowledge: Memory management, device management, systemd, containers, networking, kernel tuning
• Infrastructure as code experience: Declarative infrastructure and configuration management tools. You've managed IAM, CI runners, secrets, and machine images programmatically
• Experience with real hardware: Robot arms, depth cameras, grippers, force/torque sensors, pneumatics, or similar
• CI/CD ownership: You've not just used CI — you've owned it. Runner infrastructure, flaky test triage, build caching, GPU-enabled pipelines
• Comfort with AI coding agents: You've used tools like Claude Code, Cursor, Copilot Workspace, or similar to do real engineering work — not just autocomplete, but directing agents through multi-step debugging, refactoring, and infrastructure tasks. You understand their failure modes and know when to trust vs. verify
• Strong debugging instincts: You can go from a vague production symptom to root cause across hardware, OS, network, and application layers
• Bias toward shipping over perfecting: You fix, monitor, iterate. Your commit history has more fix: than feat: and you're proud of that

Nice to Have

• NixOS or reproducible build system experience
• Experience building or operating autonomous engineering agents/bots
• Robotics simulation (kinematics, camera models, physics)
• gRPC / Protocol Buffers
• Managed network infrastructure, VPNs, overlay networks
• Time-series databases and observability stacks

About the Work Style

This is a high-autonomy, high-output role. On a typical day you might direct an AI agent to triage overnight CI failures while you debug a production robot issue, then spend the afternoon migrating a package to a new build system. You'll write a lot of code, but you'll also write a lot of prompts — and the best candidates will see those as the same skill.