Artificial Intelligence Jobs

• London
• Hybrid · 3 days in office
• Mid-level

Join the model performance and scaling team to implement and optimize LLM post-training methods on frontier models, collaborating with research teams to turn new methods into production-ready systems. The role focuses on distributed training and inference performance, low-precision methods, and frameworks like JAX or PyTorch, in a hybrid London setting.

JAX · PyTorch · NCCL +5

• London
• Hybrid · 3 days in office
• Lead

Owns the safety assessment strategy for drug discovery projects from inception to regulatory submission, designing screening cascades and mechanistic experiments for small molecules and biologics. Guides development of an AI predictive toxicology platform with ML researchers and manages CRO relationships, as all wet-lab work is outsourced.

AI/ML

• London
• On-site
• £260,000 – £630,000 pa
• Mid-level

Owns critical parts of a production pretraining pipeline, covering performance optimization, observability, reliability and hardware debugging across the full stack. Splits time roughly 50/50 between research and engineering, designing experiments to cut step time and boost uptime, and responding to on-call launch incidents. Works in-office in London alongside SF and specialist teams.

JAX · TPU · PyTorch +3

• London
• Hybrid
• £375,000 – £640,000 pa
• Mid-level

Designs and runs large-scale reinforcement learning experiments to understand how RL behaves as model size, compute and task horizon scale, then turns validated findings into production training recipes. Builds long-horizon benchmarks and debugs failures that only surface at the research/infrastructure boundary, working closely with adjacent RL teams.

• United Kingdom
• Remote
• Permanent

Owns the path from research checkpoint to production serving infrastructure, optimising inference latency, throughput and cost using quantization, distillation, KV-cache optimisation, batching and custom kernels. Builds high-performance serving systems for real-time streaming workloads and tooling that lets researchers ship models quickly and safely.

vLLM · SGLang · CUDA +2

• United Kingdom
• Remote
• Mid-level
• Permanent

Owns large-scale distributed web crawling systems that discover, fetch and extract data from billions of pages to feed frontier AI models. Work spans content extraction from messy HTML, web-scale deduplication, recrawl and freshness strategies, politeness and rate limiting, plus tooling for researchers to request and explore crawled data.

Kubernetes

• Barcelona, Spain

Design, train and deploy reinforcement-learning controllers for robots operating in complex, unstructured real-world environments, owning the full pipeline from physics-based simulation to hardware deployment. Work spans sim-to-real transfer, domain randomisation, vision-based perception and sensor fusion, combining learned policies with classical control under tight hardware constraints.

MuJoCo · Isaac Gym · Isaac Lab +5

• London
• Hybrid
• Senior
• Permanent

Lead original technical research in AI safety, steering projects from research question to publication. Work spans interpretability, steering vectors, goal drift, machine unlearning and uncertainty calibration for transformer LLMs and agentic systems, with outputs informing frontier labs and government institutes. Involves building reproducible research code in PyTorch and Python.

PyTorch · Transformers · Metaview

• London
• Hybrid
• Lead · manages people
• Permanent

Leads a small, high-agency research team setting the scientific agenda for AI safety, with a focus on large language models and safety-critical systems. Involves conducting and overseeing red-teaming and evaluations in high-risk domains such as CBRN and cybersecurity, publishing at top conferences, and mentoring researchers while aligning work with commercial goals.

Python

• Oxford
• On-site
• £39,424 – £47,779 pa
• Entry level
• Contract

Develops novel statistical and machine learning methods for large-scale genomic, phenotypic and health data, covering areas such as genomic language models, multimodal and federated learning, and multi-ancestry methods. Builds scalable, reproducible research software applied to biobank-scale datasets, evaluates statistical and predictive performance, publishes findings and guides junior researchers in the Palamara Group.

• Oxford
• On-site
• £40,178 – £48,700 pa
• Entry level
• Contract

Designs and runs real-time in-situ X-ray imaging solidification experiments on steel using laboratory and synchrotron sources, developing multi-modal imaging methods and post-solidification microscopy. Works in the Process Dynamics group with STFC, Loughborough University and industrial partners, building bespoke experimental rigs and Python computer-vision algorithms for automated data analysis.

Synchrotron · Python

• Munich - Berlin - London - Paris

Build and scale a GPU-based simulator that trains reinforcement-learning agents for air autonomy, modelling flight dynamics and sensing across thousands of parallel environments. Work spans sim-to-real transfer, domain randomisation and RL training tooling, at the intersection of AI, GPU computing and physics-based simulation, with room to grow into leadership.

CUDA · Taichi · NVIDIA Warp +3

• London
• Hybrid
• £145,000 pa
• Director

Expression of interest for a Research Director role at the AI Security Institute, scoped at Level 7, focused on understanding advanced AI risks and translating that into government action. The position involves shaping public-interest research on frontier AI safety, with pre-release access to frontier models, ample compute, and collaboration with national security, policy and AI research experts.

• Cambridge
• Hybrid
• Permanent

Join a small, dynamic R&D team prototyping and implementing new approaches to AI problems, evaluating output quality, efficiency and feasibility against other techniques in a data-driven way. Work spans data science on large datasets, modelling and evaluation, using Python ML frameworks such as PyTorch, TensorFlow and Scikit-Learn. Open to graduates as well as senior candidates with published work in NLP, Computer Vision or Deep Learning.

Python · PyTorch · TensorFlow +1

• London
• £36,000 – £38,000 pa
• Entry level

Support senior researchers on AI projects, handling data collection, algorithm development, experimental design and analysis, and contributing to reports and publications. Includes a structured 3-month training and induction programme with mentorship, aimed at recent graduates with Python, PyTorch or TensorFlow, NumPy, Pandas and Git exposure.

PyTorch · TensorFlow · NumPy +3