Machine Learning Engineer - LLM post-training/mid-training

EPM Scientific; Southampton; On-site; GBP 70,000 - 90,000; Full time

Job summary

A pioneering technology firm seeks a Machine Learning Research Engineer to design novel approaches for model adaptation and reasoning, specifically focused on LLMs. The ideal candidate will have a PhD or equivalent experience in a related field and practical experience with LLM training pipelines. Responsibilities include exploring improvement techniques, developing reasoning paradigms, and building robust evaluation frameworks. Join a team that is redefining the boundaries of autonomous research.

Qualifications

• PhD or equivalent experience in Computer Science, Machine Learning, or a closely related discipline.
• Practical experience working with LLM training pipelines, including pre-training, mid-training, or post-training stages.
• Strong understanding of transformer architectures, optimization techniques, and representation learning principles.
• Proficiency in Python and familiarity with major ML frameworks.

Responsibilities

• Design and implement novel approaches for model adaptation and reasoning.
• Explore techniques that improve generalization, controllability, and scientific understanding.
• Develop new reasoning paradigms such as retrieval-augmented and tool-augmented approaches.

Skills

Machine Learning

Python

Transformer architectures

LLM training pipelines

Experimental validation

Education

PhD or equivalent experience in Computer Science, Machine Learning, or related discipline

Tools

PyTorch

DeepSpeed

JAX

Job description

Machine Learning Research Engineer - LLM post-training/mid-training

Our team is partnered with a materials discovery stealth venture based in San Francisco and London, led by former Oxford and Isomorphic Labs leaders in AI and experimental science. The team is pioneering large-scale language models that reason, adapt, and accelerate discovery workflows. They are combining experimental validation, synthetic data generation, and scalable infrastructure to push the boundaries of autonomous research. This is a rare chance to be part of the founding team, shaping technical direction and building systems that redefine how science is done.

The Role

• You will design and implement novel approaches for model adaptation and reasoning, exploring techniques that improve generalization, controllability, and scientific understanding.
• This includes mid-training strategies, post-training alignment, and inference-time optimization for complex workflows.
• You'll also develop new reasoning paradigms such as retrieval-augmented and tool-augmented approaches and build robust evaluation frameworks for applied scientific contexts.

Qualifications:

• PhD. or equivalent experience in Computer Science, Machine Learning, or a closely related discipline
• Practical experience working with LLM training pipelines, including pre-training, mid-training, or post-training stages
• Strong grasp of transformer architectures, optimization techniques, and representation learning principles
• Proficiency in Python and familiarity with major ML frameworks such as PyTorch, DeepSpeed, or JAX
• Knowledge of alignment and reasoning strategies, including in-context learning, chain-of-thought, tool integration, or retrieval-augmented approaches
• Ability to combine innovative research thinking with pragmatic engineering to deliver scalable, high-performance systems