The roleYou'll bridge research and production, taking ideas and turning them into systems that run at scale, stay reliable, and get better over time. Full-stack ML ownership: from raw data to deployed model.

• Day to day that looks like:Building end-to-end pipelines across data, training, evaluation, and inference
• Adapting and fine-tuning models with modern techniques: LoRA, QLoRA, SFT, DPO, distillation
• Architecting inference systems that hold up under real latency and cost constraints
• Creating data pipelines that produce high-quality synthetic and real-world training data
• Running evaluation that goes beyond benchmarks: robustness, safety, bias, production behaviour
• Owning deployment: GPU optimisation, quantisation, memory efficiency, scaling
• Working directly with application engineers so ML integrates cleanly into backend, mobile, and desktop

• Your skills and experienceDeep understanding of deep learning and transformer architectures
• Proven experience training, fine-tuning, or shipping large-scale models in production
• Strong with at least one major ML framework (PyTorch, JAX) and quick to pick up others
• Familiar with distributed training and inference tooling: DeepSpeed, FSDP, Megatron, ZeRO, Ray
• Engineering discipline: code that's readable, robust, and maintainable
• Experience optimising for GPU constraints: quantisation, mixed precision, memory
• Comfortable taking ownership of ambiguous problems from zero to one
• Ships, iterates, learns from production
• Nice to haveLLM inference frameworks: vLLM, TensorRT-LLM, FasterTransformer
• RLHF: PPO, DPO, ORPO
• Open-source contributions to ML or systems libraries
• Scientific computing, compiler, or GPU kernel experience
• Multimodal or diffusion model background
• Large-scale data processing: Arrow, Spark, Ray