Research Engineer, Human Influence — AI Safety & RL Systems

AI Security Institute (AISI); Greater London; Hybrid; GBP 65,000 - 145,000; Full time

Benefits offered by this job

Hybrid working

25 days annual leave

Learning & development stipends

Employer pension contribution

Job summary

The AI Security Institute invites applications for a Research Engineer to advance human-influence research and scalable ML evaluation pipelines. You will work across RL, LLM post‑training, and evaluation infrastructure within a hybrid setup in London, collaborating with engineers, researchers, and policy experts.

We seek candidates with strong Python production experience, ML fundamentals, and a track record of deploying large-scale evaluations or safety pipelines.

Qualifications

• Proven experience fine‑tuning or post‑training LLMs using standard libraries (PyTorch, Keras, JAX).
• Experience with RL environments and reward-based ML methods for post‑training or fine-tuning.
• Writing scalable production Python code.
• Experience serving, scaling, and containerising ML code (Docker, Kubernetes, Ray, FastAPI, SLURM) on large compute clusters.

Responsibilities

• Designing and implementing scalable RL environments for safety-focused evaluation.
• Applying interpretability methods to analyze model behavior and mitigations.
• Building the architecture for repeatable model evaluation and benchmarking.

Skills

Python

PyTorch

Keras

JAX

Reinforcement Learning

Good communication

Tools

Docker

Kubernetes

Ray

FastAPI

SLURM

Job description

The AI Security Institute invites applications for a Research Engineer to advance human-influence research and scalable ML evaluation pipelines. You will work across RL, LLM post‑training, and evaluation infrastructure within a hybrid setup in London, collaborating with engineers, researchers, and policy experts.

We seek candidates with strong Python production experience, ML fundamentals, and a track record of deploying large-scale evaluations or safety pipelines.