Autonomy Software Engineer
Every OAS company we track is hiring autonomy engineers. The bar is high — control theory + modern ML + embedded software — but so is the comp, and the mission set is genuinely interesting.
Courses for this role
Core foundations
Controls, linear algebra, probabilistic robotics. Skip these and every downstream bug looks like a mystery.
Real-time control loops don’t live in Python.
The textbook — SLAM, particle filters, everything.
Flight stacks in practice
Actually running code on something that flies. Simulators first, then real hardware.
The two open flight stacks every autonomy team starts from.
You ship maybe 2% of the hours on real hardware. The other 98% are simulated.
If your code flies commercially, you’ll meet BVLOS rules within 6 months.
Behavior trees are the dominant high-level autonomy framework on production stacks (Anduril Lattice, Shield AI Hivemind). State machines do not scale to multi-vehicle mission logic; behavior trees do.
ROS 2 is the de facto middleware for multi-vehicle autonomy.
Mission-grade
What separates a hobbyist from a hireable autonomy engineer: resilience to GPS denial, comms loss, adversarial conditions.
GPS-denied is the default assumption for defense missions.
Perception layer standardized on CNNs — worth knowing end-to-end.
Matters for any customer who requires safety-of-flight assurances.
Most airborne inference lives on Jetson. Vendor cert looks good on the resume.