Location: London (Hybrid – 1-2 days/week in office)
Reports to: Head of AI

Key Responsibilities

• Train and fine-tune large-scale foundation models for financial applications
• Scale training across multi-GPU/TPU clusters with efficient parallelization
• Manage compute resources and optimise training costs across cloud providers
• Build efficient inference pipelines for production deployment
• Implement distributed training strategies (data/model/pipeline parallelism)
• Develop custom JAX/PyTorch implementations for novel architectures
• Monitor and debug large-scale training runs with experiment tracking
• Experience training models with 1B+ parameters
• Strong expertise in JAX and/or PyTorch at scale
• Hands-on experience with multi-GPU/TPU training and optimization
• Deep understanding of diffusion models and autoregressive architectures
• Experience with distributed training frameworks (DeepSpeed, FSDP, Accelerate)
• Experience with foundation model fine-tuning (LoRA, QLoRA, PEFT)
• Strong understanding of attention mechanisms and transformer architectures

Nice to Have

• Experience with mixture-of-experts (MoE) architectures
• Knowledge of quantization and model compression techniques
• Experience with streaming/online learning at scale
• Background in financial modeling or time series
• Familiarity with MLOps tools (Weights & Biases, MLflow)

What We Offer

• Work on cutting-edge AI for major financial institutions (LSEG, Barclays)
• Access to significant compute resources (GPUs/TPUs)