In short
12-month hybrid contract in London for a leading AI research organisation: build agent frameworks, LLM training/fine-tuning pipelines, evaluation benchmarks and large-scale experiment infrastructure that let AI agents run research autonomously. Core stack: Python, PyTorch, distributed GPU training (FSDP/DDP) and containerised code-execution environments.
About the role
A global technology organisation is looking for an AI Research Engineer to develop AI agents that support researchers in planning experiments, generating and executing code, evaluating outcomes, and identifying the next best steps in the research process.
Details:
Contract length: 12 months
Location: London, UK
Model: Hybrid
Description
A leading AI research organisation is developing advanced AI agents designed to accelerate the research process itself. These systems are capable of generating hypotheses, designing and executing experiments, analysing results, and iteratively improving their approach with minimal human intervention.
The team focuses on building autonomous research systems that can tackle complex technical tasks, including machine learning engineering, coding, experimentation, and scientific discovery. Their work spans agent architecture, model training and evaluation, benchmarking, large-scale experimentation infrastructure, and distributed computing systems. By combining cutting-edge AI research with robust engineering, they aim to create tools that significantly increase the speed, quality, and scalability of AI innovation.
The Role
Research Engineers partner closely with Research Scientists to transform novel research ideas into reliable, scalable, and reproducible systems. You will take ownership of well-defined components of the research platform, delivering solutions from concept through deployment.
You could be responsible for developing benchmarks, agent capabilities, evaluation frameworks, training workflows, experimentation infrastructure, or research tooling. Success in this role requires strong software engineering fundamentals, hands-on experience with modern AI systems, and the ability to independently drive projects that support cutting-edge research.
Responsibilities
- Design, build and maintain agent frameworks, tooling and environments that enable AI agents to perform research-focused tasks such as machine learning engineering, software development and experiment execution.
- Build, launch and manage large-scale experiments across distributed GPU infrastructure, monitoring progress and analysing outcomes.
- Develop evaluation frameworks and benchmarks, including automated scoring systems, assessment rubrics and reproducibility validation.
- Implement, optimise and extend training, fine-tuning and post-training workflows for large language model-based agents.
- Improve the reliability, efficiency and cost-effectiveness of large-scale agent execution and compute utilisation.
- Convert research prototypes into reusable, production-quality libraries, tools and services.
- Contribute to research papers, technical publications, open-source projects and internal knowledge sharing.
- Participate in operational support rotations focused on experimentation infrastructure, evaluation systems and reproducibility.
Minimum Qualifications
- Bachelor's or Master's degree in Computer Science, Engineering, Mathematics or a related field, or equivalent practical experience.
- Software engineering or research engineering experience, or a PhD with significant engineering contributions.
- Strong Python programming skills and experience with PyTorch or a comparable deep learning framework.
- Hands-on experience training, fine-tuning or evaluating large language models.
- Experience running distributed workloads and large-scale experiments on GPU clusters.
- Strong software engineering practices including testing, code reviews, version control and modular system design.
Preferred Qualifications
- Experience developing AI agents that utilise tools, planning capabilities, multi-step reasoning or code execution environments.
- Familiarity with benchmarking frameworks used to evaluate coding, machine learning or research-oriented AI systems.
- Experience with reinforcement learning for language models or other advanced post-training techniques.
- Experience designing robust evaluation methodologies, including automated judgement systems, rubric-based assessments and statistically rigorous analysis.
- Publications at leading AI conferences such as NeurIPS, ICML, ICLR or ACL, or significant open-source contributions.
- Experience collaborating with academic researchers and cross-functional technical teams.
Technical Skills
- Expert-level Python development, including Linux and Bash-based workflows.
- PyTorch, distributed training technologies (FSDP, DDP) and inference serving platforms.
- Large language model fine-tuning, reinforcement learning and post-training pipelines.
- Agent frameworks, tool and function calling, prompt engineering and agent architecture design.
- Sandboxed code execution environments, containerisation technologies and job schedulers.
- Experiment tracking, data pipelines and large-scale analysis workflows.
- Benchmark development, evaluation design, metrics creation and statistical testing.
- GPU performance optimisation, profiling and infrastructure scaling.
- Reproducibility practices, including experiment tracking, configuration management and artifact versioning.
Research Skills
- Reading and rapidly implementing cutting-edge research papers.
- Designing ablation studies and controlled experimental comparisons.
- Analysing ambiguous or noisy experimental results and identifying actionable next steps.
- Producing clear technical documentation, reports and research publications.
Collaboration & Execution
- Own medium-sized technical projects from planning through delivery with minimal supervision.
- Translate ambiguous research objectives into structured engineering tasks and milestones.
- Communicate effectively with researchers, engineers and cross-functional stakeholders.
- Support and mentor junior engineers and interns through collaboration and code reviews.
- Provide realistic delivery estimates, proactively identify dependencies and raise risks early.
Please apply for more information.