What You'll Do

• Design and implement end-to-end LLM training pipelines
• Source and, where appropriate, preprocess datasets for training and evaluation
• Fine-tune and optimise open weight models (LLMs, vision or traditional ML)
• Build evaluation frameworks and define performance metrics
• Develop and maintain data pipelines and training workflows
• Analyse training pipelines and optimise for latency, cost and scalability
• Implement monitoring, logging and feedback loops for continuous improvement
• Experiment with modern AI tooling and services to assess how they can be leveraged

What You'll Bring

• Proven experience training and fine-tuning LLMs or multimodal models (not just consuming APIs)
• A solid understanding of model evaluation and validation, overfitting and bias/variance trade-offs, and data quality and feature engineering
• Proficiency in Python and ML frameworks such as PyTorch or TensorFlow
• Experience building and maintaining ML pipelines in production
• Familiarity with GPU usage and optimisation
• The ability to debug and improve model performance systematically

Also Valued

• Knowledge of distributed training or large-scale data processing
• Experience with MLOps tools (CI/CD for ML, experiment tracking, model versioning)
• A background in applied research or publishing
• Familiarity with retrieval systems, embeddings or ranking models

Ideally you'll have a maths or computer science research background with a focus on developing new algorithms or techniques for training and deploying AI models, whether gained in a large organisation, a start-up, or academia, with an emphasis on cutting-edge machine learning.