Tübingen or Munich · Full-time
Simulation Engineer
| Optics
Develop robust, high-performance physics simulators for optical systems. You will turn wave propagation and optical-system behavior into accurate, differentiable JAX programs that run efficiently on GPUs and scale from rapid experiments to large simulation campaigns.
The role
At Feyer, physics simulators are the environments in which our AI systems learn what is physically possible. Their speed, accuracy, and reliability determine the scale and quality of every experiment we can run.
In this role, you will own core parts of the optical simulation stack, from modeling assumptions and numerical methods to GPU implementation, validation, and performance. You will work closely with colleagues in physics, AI, and engineering while remaining deeply focused on building exceptional simulation software.
What you will work on
- Own optical simulation methods: design and implement numerical models of propagation, illumination, interference, and optical-system response.
- Run simulations at scale: build differentiable simulation pipelines in JAX and optimize memory use, throughput, batching, and numerical precision on GPUs.
- Establish trustworthy validation: quantify convergence, stability, and physical accuracy against analytical references, established solvers, and experimental data.
- Model real systems: capture aberrations, tolerances, calibration drift, noise, and other effects that determine whether a simulation remains predictive.
- Build durable scientific software: create modular APIs, benchmarks, tests, and documentation that make the simulator easy to extend and hard to misuse.
What you bring
- A strong foundation in physics, applied mathematics, computer science, or a related engineering field.
- Hands-on experience with numerical methods, computational physics, scientific software, or adjacent simulation work.
- Knowledge of Fourier or wave optics, electromagnetic propagation, physical optics, or a closely related field.
- A careful approach to verification, numerical testing, and building reliable simulation software.
Nice to have
- Experience with differentiable simulation, adjoint methods, inverse design, or gradient-based optical optimization.
- Experience with photonic systems, optical-system modeling, or experimental optics.
- Experience designing and profiling efficient JAX programs using automatic differentiation and transformations such as jit, vmap, and multi-device sharding.
- Experience scaling simulation workloads across multiple GPUs or compute nodes.
What we offer
- Own a simulation stack central to Feyer's AI systems, with the freedom to shape its numerical methods and software architecture.
- Access to hundreds of thousands of GPU-hours for large-scale simulation and AI experiments.
- One of 10 teams selected from 500+ applicants for SPRIND's Next Frontier AI Challenge, with a path to up to €26.5 million in equity-free funding per team.
- Work directly with an experienced founding team in a small interdisciplinary group spanning physics, AI, and engineering.
- A competitive salary that reflects your experience and impact.