About the role
About us
AIM Solutions is an early-stage deep-tech startup in Eindhoven building inspection and metrology equipment for the semiconductor industry. We are in active build phase, small, and growing.
The role
You will own the algorithms at the core of our instrument: the GPU-accelerated engine for phase retrieval and wafer metrology. This is the deep end of computational imaging, where progress comes from understanding both the optimization and the physics.
Concretely, you will:
- Develop and maintain our GPU-accelerated computational imaging engine for phase retrieval
- Improve reconstruction quality, convergence, and robustness across varied and imperfect datasets
- Design and implement new reconstruction methods: physics-based differentiable forward models, regularization and constraint schemes for ill-posed problems, and support for new acquisition modes
- Translate methods from the literature and from our own ideas into tested, performant code
- Use physical insight into the optical system to decide how to change the algorithm so that results improve
- Profile and optimize the engine for GPU performance on large datasets
Your profile
- A PhD, or equivalent research depth from industry, in a physics-heavy or numerically heavy field
- Several years developing numerical or scientific algorithms, ideally in imaging, inverse problems, or a comparable domain
- Strong numerical optimization, both gradient-based and iterative methods
- Working knowledge of Fourier optics and wave-propagation physics
- Strong Python and array programming, with experience in GPU frameworks such as JAX, PyTorch, TensorFlow, or CUDA
- You write tested, maintainable code that others can build on
- You can reason from physics to algorithm and back, and judge what will improve a reconstruction
- Technical curiosity; we will teach you the specifics of our computational imaging approach
- Full professional proficiency in English
Bonus experience in any of these
- Phase retrieval, coherent diffractive imaging, holography, or tomography
- Physics-informed machine learning, differentiable forward models, automatic differentiation
- Regularization and constrained optimization for ill-posed inverse problems
- High-performance and GPU computing with performance profiling
- Linux development environments
This is a senior, research-to-production role. We expect a track record of taking new methods from concept to working, maintained code that runs in a real instrument.
What we offer
- Ownership of the reconstruction algorithms at the heart of a tool that does not exist anywhere else yet
- Small team, short feedback loops, an open environment to have real impact
- Competitive salary for the Dutch market