Senior Autonomy Software Engineer

Our client is an early-stage defense technology startup building autonomous vehicles for the U.S. military. They're already under active government contract, their first vehicle heads to military range testing in the coming months, and it's slated for real operational use shortly after. If you join now, the autonomy you write is live in months, not years.

The team is lean, fast and deeply technical. Leadership includes a former lead powertrain engineer on one of the highest-performance production EVs ever built and a career Navy officer who ran unmanned programs. The head of autonomy has stood up flight software stacks from scratch at some of the most closely watched hard-tech startups of the last few years. You'll also work alongside a senior AI leader from one of the world's largest tech companies, who advises the team part-time.

Their long-term play is Tesla- and SpaceX-style vertical integration: owning the critical systems in-house instead of buying them. For autonomy, that means no inherited codebase and no off-the-shelf robotics framework holding the vehicle together. You design the architecture.

They're growing from roughly 15 people today toward 5x plus over the next 12 months, and they intend to promote from within as the team scales.

This is a direct hire, permanent placement position for a client of AdAstra Talent Advisors.

About the Role

This is the first dedicated autonomy hire, and a Responsible Engineer role modeled on the SpaceX approach: you own the autonomy and perception stack, you make the architectural calls, and the buck stops with you (with real peer review along the way). It's hands-on IC work. The head of autonomy is handing off his own IC work and wants you to tell him how the system should be built.

The expectation is clear: stand up the vehicle's autonomy architecture from a blank sheet and get a working product running in about 90 days. You won't write every component from zero. Proven open-source tools (YOLO, OpenCV and the like) come in where they make sense. But there's no ROS; the glue, the middleware and the overall architecture are yours to write.

The stack spans three areas, and the first hire needs to be strong across all of them:

• Perception. Detect, recognize, classify and track what the vehicle's sensors see, using machine learning and classical computer vision, on constrained edge compute.
• Navigation and localization. Know where the vehicle is without GPS: state estimation, sensor fusion and mapping.
• Planning and decision-making. Mission planning, path planning, obstacle avoidance, and recovering safely when something goes wrong.

The team plans to add two or three more autonomy specialists behind this hire (perception, and mission planning/navigation), so deep specialists are encouraged to apply.

Responsibilities

What You'll Own

• Requirements, architecture and the hard tradeoffs, often with incomplete information
• Production autonomy software in C, C++ or Rust, running within real compute, memory, latency and power limits
• Model training, evaluation and deployment for perception
• The simulation, replay, SIL and HIL environments used to test autonomy before it goes in the water
• Sensor selection and sensing architecture, alongside avionics and embedded software
• Field testing and sea trials, and root-cause analysis from logs and telemetry afterward
• Mentoring the autonomy engineers who join after you

Who They're Looking For

Engineers who have built an autonomy system from a blank page and shipped it to a real vehicle, ideally more than once. You can name the filter, the planner, the model and why you chose it. You've owned the decision, not just implemented it.

• Deep, practical experience in perception, navigation or planning for real-world autonomous systems
• A track record of building autonomy from scratch, not extending a mature stack
• Production code in a systems language, not just prototypes, notebooks or papers
• Comfort writing your own architecture and glue instead of relying on a framework
• Smart, skeptical use of AI coding tools to move fast without shipping slop
• First-principles thinking and a bias for speed

What Success Looks Like

• In about 90 days, the vehicle runs on an autonomy architecture you designed and can defend
• Your perception, navigation and planning are running in the water during range testing, and you're the one reading the logs
• The test and replay tooling you built lets the team ship changes with confidence
• As the autonomy team grows, you're a natural candidate to lead it

Requirements

• Relevant experience with hands-on autonomy, perception, robotics or production