Physical AI Engineer - SW

Archer is developing electric vertical-takeoff aircraft, and our SW team builds the advanced simulation, machine learning, and engineering tooling that supports how those aircraft are designed and analyzed. We are looking for a Physical AI Engineer who works at the intersection of scientific machine learning, software engineering, and aerospace — building learned models of physical systems and the AI-driven workflows that put them to work.

This is a hands-on research-and-build role. You will train models that approximate expensive physics, integrate foundation models into engineering tooling, and turn promising research into reliable, well-tested software that other engineers depend on.

• Build, train, and validate machine-learning models that approximate the behavior of physical systems — neural operators, physics-informed networks, and related surrogate models — to evaluate engineering questions far faster than traditional simulation, with calibrated, honest uncertainty.
• Generate and curate large-scale synthetic datasets — parametric geometry paired with high-fidelity physics solves — to train and stress-test those models.
• Build learned models that work alongside traditional CFD/FEA and optimization solvers, so engineers get fast answers without giving up trusted ones.
• Integrate frontier foundation models (e.g., Claude) into agentic engineering workflows, where the model orchestrates, routes, and drafts — and verified computation plus human judgment govern the outcome.
• Build ML systems whose outputs are reliable and traceable, so the results engineers act on can be trusted and checked.
• Take research from paper or prototype to production: ship into a typed, tested Python monorepo with real reproducibility — not one-off notebooks.
• Partner with aerodynamics, structures, propulsion, GN&C, and avionics engineers to turn their analyses into automated, dependable workflows.
• Help connect simulation to reality — comparing model predictions against test-rig and flight data and improving the models from what you learn.

• Strong programming fundamentals and excellent Python, with a track record of building and scaling ML or data pipelines inside a real, version-controlled codebase — and the testing discipline and reproducibility that production systems require.
• Hands-on machine learning experience: training, evaluating, and debugging models, and a demonstrated ability to take a research idea to a working, tested implementation.
• Working knowledge of scientific machine learning — physics-informed models, neural operators, or surrogate modeling — or a strong applied-math, numerical-methods, or simulation background and the ability to ramp into it quickly.
• Experience generating or working with synthetic data to train learned systems.
• Sound judgment about foundation models: you have integrated them into software, and you understand where a model can be trusted and where it must be backed by verified computation or a human decision.
• An evidence-first instinct — you treat a model's output as only as good as the data and verification behind it, and you build systems that make that explicit.
• BSc, MSc, or equivalent experience in a quantitative or engineering discipline (computer science, applied math, mechanical/aerospace engineering, physics, or related).
• Solid command of Git and modern software-development best practices.
• Strong communication and the ability to collaborate across software, hardware, and engineering disciplines.
• Genuine interest in aviation and in building learning systems that hold up under real-world scrutiny.

• Background in aerospace, mechanical, or a physical-sciences domain; familiarity with CFD, FEA, or multidisciplinary design analysis and optimization (MDAO).
• Experience with differentiable optimization, constrained learning, or enforcing physical constraints inside learned models.
• Exposure to safety-critical or other regulated-systems environments — or a real appetite to learn how they work.
• Sim-to-real techniques (domain randomization, system identification) and experience reconciling models against hardware or flight-test data.
• Hands-on lab instrumentation (oscilloscopes, logic analyzers, protocol analyzers, HIL/SIL rigs) — valuable where the work meets real test hardware.
• Fluency in the modern scientific-Python and ML-systems stack (PyTorch/JAX, async services, job queues, vector or time-series databases).
• Understanding of model-scaling principles and their practical trade-offs.

You will do real research and apply it to a product: rigorous, tested, and trustworthy, because people will fly behind it.