Senior Deep Learning Research Scientist

Senior Research Scientist, Geometric Deep Learning and Scientific ML

Location: Berlin, Germany

Working model: Full time, hybrid

About the company

We are supporting an early stage deep learning company developing foundation models for engineering and physical systems.

The company works with industrial partners across mechanical engineering, electrical engineering, manufacturing, and engineering design. Its goal is to create models that can generalise across different physical problems, geometries, and industrial constraints.

This is a small research led team where new ideas can be implemented, tested, and improved quickly.

The opportunity

We are looking for a Senior Research Scientist who can develop original deep learning methods for complex engineering problems.

This is not a role focused on applying existing models without questioning them. You will be expected to understand why an architecture works, identify where current methods fail, and develop new approaches from first principles.

You should have deep research expertise in at least one relevant area, while being willing to work across adjacent fields as projects develop.

What you could work on

• Designing and validating new geometric deep learning architectures
• Developing models for graphs, meshes, point clouds, particles, surfaces, and other structured data
• Building generative models for engineering design and physical systems
• Developing surrogate models for computationally expensive simulations
• Training models using synthetic and simulated data
• Conditioning generative models on multimodal inputs and physical constraints
• Exploring neural operators, neural differential equations, diffusion models, and function space modelling
• Studying network architecture, optimiser behaviour, regularisation, loss geometry, and generalisation
• Translating research papers into reliable experimental systems
• Working directly with industrial simulation environments and proprietary engineering datasets

What we are looking for

• A PhD in computer science, mathematics, applied mathematics, physics, engineering, or a related subject
• Research depth in geometric deep learning, scientific machine learning, generative modelling, synthetic data, surrogate modelling, neural operators, or a closely related area
• Strong mathematical understanding of neural networks, optimisation, and generalisation
• Evidence that you can develop original methods rather than only reproduce existing research
• Experience designing, training, and evaluating deep learning architectures
• Ability to read papers critically and explain the reasoning behind technical decisions
• Strong Python experience with PyTorch, JAX, TensorFlow, or similar frameworks
• Interest in applying research to difficult industrial problems
• Candidates completing a strong PhD, experienced postdoctoral researchers, and researchers with several years of industry experience are all encouraged to apply.

Useful additional experience

• Equivariant neural networks
• Graph neural networks
• Three dimensional geometry, meshes, point clouds, or particle systems
• Physics informed learning and differentiable simulation
• Synthetic data generation
• Reinforcement learning for optimisation or trajectory generation
• Procedural geometry generation
• Simulated data transfer into real applications
• Experience moving research models into production environments
• Previous CAD, CAM, CNC, or manufacturing experience is not required.

What is offered

• Direct influence over the company’s research direction
• Close collaboration with the founders
• Freedom to explore and test original ideas
• Access to industrial datasets and simulation environments
• Short research and development cycles
• Conference attendance and continued learning support
• Flexible working hours
• Potential equity participation
• Possible relocation support
• English speaking working environment, with no German requirement

The team works from Berlin, with approximately two to three days per week available to work from home.

If you enjoy developing new deep learning methods, reasoning from mathematical foundations, and applying research to physical engineering problems, we would be interested in speaking with you.