Senior Machine Learning Software Engineer, Research

About us Re-architecting Engineering for the Age of Intelligence PhysicsX is the physics AI company for industrials. The company’s mission is to accelerate hardware innovation by overhauling what industrial engineering and manufacturing look like today. PhysicsX is building a new simulation software stack to deliver deep physics AI enablement across the entire engineering lifecycle. The company partners with leading organisations in aerospace & defence, automotive, semiconductors, materials, and energy & renewables, supporting them on some of their most critical and complex challenges. PhysicsX is headquartered in the United Kingdom, with offices in London, New York, and Singapore and an expanding presence in the Bay Area. Note: We are currently recruiting for multiple positions across different levels, however please only apply for the role that best aligns with your skillset and career goals. What you will do • Own the training stack end to end: experiment harness, sweeps, distributed training, checkpointing, and throughput. • Implement and optimise model architectures with efficiency perspective: sparse attention and convolution, custom kernels, memory and precision work, and making a prototype fast enough to train at scale. • Take architectural ideas from prototype to trained model, and own the ablations that establish whether a change actually helped. • Set technical direction for the training stack, and build the inference and evaluation path that turns a checkpoint into something usable. • Identify engineering investments that unlock research progress. What you bring to the table • Senior-level experience in training machine learning models at scale, in industry or a research group of comparable intensity, instrumental in: • distributed training with FSDP or equivalent, including profiling and diagnosing idle GPUs; • PyTorch and CUDA at depth, ideally including sparse convolution or sparse attention internals and custom CUDA or Triton kernels; • implementing architectures from papers and making them work, then making them fast; • reproducible experiment infrastructure, because most of the compute in a programme like this is spent on runs that fail. • Experience with 3D or other irregular data: meshes, point clouds, voxels, sparse tensors. • Ability to scope and deliver projects, and strong problem-solving skills. • Degree in CS, engineering, mathematics, physics or a related field. A PhD is welcome but not required; engineering depth matters more. • Desirable: open-source contributions, public technical writing, or papers at MLSys, SIGGRAPH or CVPR What we offer Build what actually matters Help shape an AI-native engineering company at a formative stage, tackling problems that genuinely matter for industry and society. This is work with real-world impact - and something you can be proud to stand behind. Learn alongside exceptional people Work with a high-caliber, collaborative team of engineers, scientists, and operators who care deeply about doing great work, and about helping each other get better. We come from diverse backgrounds, but we share a commitment to operating at the highest level and addressing some of the most complex challenges out there. If you’re ambitious, thoughtful, and driven by impact, you’ll feel at home. Influence over hierarchy We operate with a flat structure: good ideas win - wherever they come from. Questioning assumptions and challenging the status quo isn’t just welcomed, it’s expected. Sustainable pace, long-term ambition Building meaningful technology is a marathon, not a sprint. We believe in balancing focused, ambitious work with a life beyond it. Our hybrid model blends time together in our Shoreditch office with work-from-home days, giving you the flexibility t