As the Staff Engineer, AI Autonomy, you'll be at the forefront of bringing intelligent, learned behavior to real aircraft — building and deploying the vision-language and VLA models that let Archer's eVTOL aircraft, including Midnight, perceive, reason, and navigate in the real world. You will develop and integrate multi-model systems, design benchmarks, fine-tune foundation models, validate through data-driven testing, and take models from prototype to deployed, monitored components running on edge hardware.

Responsibilities

Develop and integrate multi-model systems; VLN/VLA, perception, and language models.
Design benchmarks, measure success and failure modes against internal baselines and published results, and report comparisons transparently.
Fine-tune and adapt foundation models, weighing tradeoffs clearly at each step.
Validate through data-driven testing in simulation, on hardware, and in flight test, partnering closely with the Safety and Simulation Engineers who check your model's output.
Take models from prototype to deployed, monitored components running on edge hardware under real latency and power constraints.

Requirements

8+ years of experience related to position minimum
M.S., PhD, or equivalent experience in Computer Science, Robotics, or a related field, with strong applied machine learning depth (exceptional BS candidates also considered).
Hands-on experience building, fine-tuning, and/or integrating vision-language or VLA models.
Solid grasp of transformer architectures and multimodal foundation models, including the practical mechanics of fine-tuning a pre-trained checkpoint to a new domain.
Strong with PyTorch (or comparable frameworks); comfortable owning model-serving integration end to end.
Real 0→1 experience — you've built the first version of something, not just iterated on a mature product.
Comfort with ambiguity. If you need a fully scoped program before you can start, this won't be a fit.

Nice to have

VLN or embodied-agent navigation experience (aerial, ground, or indoor robotics).
Experience with reinforcement learning, imitation learning, or self-supervised training for navigation or control.
Edge deployment experience (NVIDIA Jetson-class or comparable) — measured and tuned inference latency and memory footprint.
Shipped evaluation tooling or benchmarks that enable data-driven decisions.
Experience in or alongside a safety-critical or regulated engineering process (aerospace, automotive, medical devices), even outside an ML-specific role.