Machine Learning Scientist – Large Multimodal Models

This is a fully remote position, open to applicants in United Kingdom.

📋 Description

• Conduct research and devise strategies for post-training of large-scale multimodal foundation models.

• Create reward functions, training objectives, data generation strategies, and evaluation protocols for reinforcement learning and other post-training methods applied to multimodal LLMs.

• Establish systematic experimentation and hyperparameter optimization workflows to investigate post-training recipes, model configurations, and training strategies.

• Develop and implement inference optimization techniques for high-throughput model evaluation and interactive discovery workflows.

• Design and manage benchmarking and evaluation frameworks across various modalities, downstream tasks, and scientific applications.

• Collaborate with ML and software engineering teams to productionize models, evaluation systems, and inference services.

• Work alongside computational chemists, medicinal chemists, and biologists to align model development and post-training objectives with drug discovery needs.

• Present findings to internal teams, external collaborators, and at conferences.

• Write, refactor, test, document, and package high-quality research and engineering code.

• Research and create AI-driven discovery technologies for Iambic Therapeutics' drug discovery platform.

⛳️ Requirements

• PhD in machine learning, computer science, computational chemistry, physics, or a related computational STEM field, or equivalent industry experience showcasing comparable depth.

• Proficient in Python and PyTorch, including the implementation, training, debugging, and evaluation of deep learning models from start to finish.

• Proven experience in training large-scale transformer models.

• Demonstrated expertise in reinforcement learning techniques such as RLHF, RLAIF, PPO, GRPO, RL with verifiable rewards, or similar methods (strongly preferred).

• Experience with supervised fine-tuning, full-parameter fine-tuning, parameter-efficient fine-tuning (LoRA), or related techniques.

• Systematic hyperparameter optimization or large-scale experimentation utilizing tools like Optuna, Ray Tune, or comparable frameworks.

• Strong engineering practices: reproducible experimentation, clean code, thorough testing, and performance-aware debugging.

• Familiarity with contemporary ML infrastructure such as Docker, CUDA, Kubernetes, and experiment tracking tools like Weights & Biases.

• Experience with multimodal or multi-task model architectures (preferred).

• Knowledge in training and inference optimization, including mixed precision, kernel optimization, quantization, or distributed strategies (preferred).

• Familiarity with biomedical, chemical, or biological data domains (preferred).

• Experience with distributed training at scale (preferred).

• Background in HPC or large-scale training operations (preferred).

🏝️ Benefits

• Private medical insurance.

• Life assurance.

• Pension contributions.

• Flexible holiday allowances.

• Modern and collaborative work environment located in the center of Bristol.