Member of Technical Staff — Training Infrastructure
CausalOur mission is general causal intelligence; AI that is capable of (1) predicting the future and (2) identifying the actions to alter it.
To achieve this breakthrough, we are building a Large Physics foundation Model (LPM) because physical systems, unlike text or images, are governed by verifiable cause and effect. We believe that scaling on physics will enable an understanding of causality required to predict and control physical systems, starting with weather.
Our founding team has built and deployed AI against the physical world in robotics, drug discovery, and particle physics at institutions like DeepMind, Waymo, Cruise, Insitro, Nabla Bio, and CERN.
We look for infrastructure engineers who are excited to tackle unsolved problems. Training an LPM means scaling novel architectures over multimodal physical data — a problem where the playbooks from language and vision only partially apply. Your mission is to make large-scale training fast, efficient, and reliable, so that every GPU cycle accelerates research progress.
Responsibilities
Design, implement, and optimize distributed training systems that scale across thousands of GPUs
Research and test parallelization strategies and numerical precision trade-offs across model scales, including for architectures that don't map cleanly onto existing LLM training stacks
Analyze, profile, and debug low-level GPU operations to maximize throughput and hardware utilization
Build reusable frameworks for checkpointing, fault tolerance, and reproducibility that stay robust under rapid research iteration
Collaborate with researchers to bring novel model architectures from prototype to full scale
Stay up-to-date on research to bring new ideas to work
What we're looking for
We value a relentless approach to problem-solving, rapid execution, and the ability to quickly learn in unfamiliar domains.
Demonstrated proficiency with distributed training frameworks and techniques (e.g. FSDP, DeepSpeed, Megatron, Pytorch, JAX/XLA) to train large foundation models
Strong grasp of state-of-the-art techniques for optimizing training workloads: parallelism strategies, memory optimization, mixed precision, communication overlap
Ability to profile and debug performance in complex codebases, from framework internals down to kernels and collectives
Deep understanding of deep learning frameworks (e.g. PyTorch, JAX) and their underlying system architectures
Bonus: contributions to open-source ML infrastructure (e.g. PyTorch, Megatron-LM, DeepSpeed, XLA)
Similar roles
Design & Tech
Related reads from TCHNX
The Quiet Revolution in Local-First Software
As major platforms face outages and data breaches, a new generation of developers is building applications that prioritise local data storage and peer-to-peer sync, challenging the cloud-first orthodoxy that's dominated tech for two decades.
The Quiet Revolution in Edge AI: Why Your Next Computer Might Not Need the Cloud
As neural processing units become standard in consumer devices, we're witnessing a fundamental shift in how AI applications work. Local processing is no longer a fallback; it's becoming the preferred architecture.
The Rise of AI-Assisted Code Generation 2: Are Developers Becoming Prompt Engineers?
As AI coding assistants reshape software development, the industry grapples with a fundamental question: is writing code giving way to writing prompts? We examine how London's tech scene is adapting to this seismic shift.