The role is about designing and training new neural network models that solve the equations behind airflow and other physical behavior, then getting them running reliably inside a real product. A strong candidate has several years of research experience, a solid publication record, and deep skill in a framework like PyTorch or JAX. They also bring experience training models across many GPUs and the habit of shipping work rather than leaving it in the lab.
How it fits the AI era
This job sits at the center of the AI shift in engineering, since the models themselves replace slow simulation runs with much faster predictions. Tools that write code and speed up experiments can help with routine implementation, but choosing what to research, checking that results obey physical laws, and judging whether a prediction can be trusted still take human expertise. Mentoring others and setting research standards are also very human parts of the work.
How it builds AI experience
Someone in this role would gain hands-on experience building foundation-style models for physics, including geometry handling, large-scale training and fast inference for engineering customers in automotive, aerospace and energy. To build toward it, a candidate could reproduce work on operator learning or graph networks for fluid problems and test it on public flow datasets. Writing up how well the model holds up against a trusted solver shows the right mindset.
Worth showing
- Published ML research at top venues
- PyTorch or JAX custom models
- Distributed GPU training experience
- Models shipped to production
- Physics-ML such as FNO or PINNs
This is our own read, written from the employer’s posting and the company’s Spotlight. The employer’s posting is the source of truth for the role.
About BeyondMath
Foundational physics artificial intelligence platform accelerating engineering simulation and multi-physics aerodynamic design optimization.
Tools mentioned: deep learning, C++, PyTorch, TensorFlow, CUDA
#J-18808-Ljbffr…