> Markdown version of [/jobs/ext/3742692-ai-researcher](https://www.wearedevelopers.com/jobs/ext/3742692-ai-researcher). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI Researcher

- **Company:** Beyondmath
- **Location:** UK
- **Experience:** Expert
- **Contract:** Permanent contract
- **Skills:** Artificial Intelligence, Artificial Neural Networks, C++ (Programming Language), Computational Fluid Dynamics, Computer Clusters, Nvidia CUDA, Distributed Computing Environment, Machine Learning, Tensorflow, Graph Neural Networks, Pytorch, Deep Learning, Information Technology, Data Pipelines
- **Published:** October 15, 2026
- **Apply:** https://startup.jobs/ai-researcher-beyondmath-10449383

## About the Role

You are a rare hybrid: a scientist who loves the elegance of a theorem, but an engineer who gets a thrill from seeing a model successfully optimize a real-world turbine or airframe. You thrive in the ambiguity of a "greenfield" opportunity and have the grit to solve problems where no textbook solution exists., * PhD or MSc in Computer Science, Physics, Mathematics, or a related quantitative field.

 * 5+ years of post-grad experience in AI/ML research, with a demonstrable track record of models made it from the lab into production environments.
 * Deep Technical Mastery: Expert-level proficiency in PyTorch, JAX, or TensorFlow, with a focus on building custom layers, loss functions, and optimization loops.
 * Published Excellence: A strong record of high-quality publications in top-tier venues (e.g., NeurIPS, ICML, CVPR, or physics-specific AI journals).
 * Systems Thinking: Experience with scalable training infrastructure, including distributed training across GPU clusters and data pipeline automation.

Highly Desirable:

 * Physics-ML Expertise: Experience with Physics-Informed Neural Networks (PINNs), Operator Learning (DeepONet/FNO), or Equivariant Neural Networks.
 * Domain Knowledge: Familiarity with Aerodynamics, Fluid Dynamics, or Structural Mechanics.
 * Engineering Rigor: Familiarity with C++, CUDA for low-level model optimization.

## Description

We are moving beyond the "generic AI" hype to solve the world's hardest physical engineering challenges in automotive, aerospace, and energy. We are looking for a Senior AI Researcher who balances scientific curiosity with the engineering discipline required to see models thrive in production environments., As a Senior AI Researcher, you will be a core architect of our technical roadmap. This is not a "siloed" research role; you will lead the transition from theoretical breakthroughs in Physics based Deep Learning to robust, scalable systems used by world-class engineers.

You will have the creative freedom to set research agendas while ensuring our models remain grounded in physical reality and industrial-scale performance.

Key Responsibilities

 * Architect Physics-AI Foundations: Lead the research and development of novel ML architectures (e.g., Transformers, GNNs, or Diffusion models) designed specifically to solve complex partial differential equations (PDEs) including aerodynamic simulations.
 * Bridge Research & Production: Translate high-level mathematical concepts into clean, high-performance code. You won't just "throw models over the wall"; you will ensure they are optimized for inference and integrated into our production design platform.
 * Advance Geometry Representation: Pioneer new ways to represent complex geometric design variations for efficient use in deep learning models.
 * Strategic Leadership: Mentor junior researchers and engineers. Help define our internal research standards, reproducibility pipelines, and high-performance compute (HPC) infrastructure requirements.
 * External Impact: Represent BeyondMath in the global AI community. Publish influential research at top-tier conferences (NeurIPS, ICML, ICLR) and position the company as the leader in "AI for Physics."
 * Cross-Functional Collaboration: Partner with CFD specialists and software engineers to ensure our models respect physical constraints while maintaining thespeed advantages of neural networks.