Artificial Intelligence Jobs

Anthropic · London

A four-month, full-time research fellowship focused on empirical AI safety and security projects, using external infrastructure such as open-source models and public APIs, with the goal of producing a public output like a paper submission. Fellows receive direct mentorship from Anthropic researchers, access to shared workspaces in Berkeley or London, and join a broader AI safety and security research community.

PhysicsX · Singapore

Builds physical foundation models for industrial engineering, scaling training across multi-GPU cloud compute and turning research prototypes into robust, optimised implementations. Works closely with research scientists, simulation engineers and customers, owning research work-streams and translating results into reusable libraries, tooling and products.

PhysicsX · London

Builds and owns high-fidelity computational plasma models for semiconductor processes such as etching, CVD/ALD deposition and ion-driven surface interactions, using tools like COMSOL, HPEM, VizGlow and CFD-ACE+. Works directly with semiconductor customers, often on-site, to scope challenges and validate simulations, and collaborates with data scientists and ML engineers to embed the work into AI-powered workflows.

CrowdStrike · United Kingdom

Join the Advanced Research Team to conduct vulnerability research and exploit development, analysing vendor patches to identify vulnerabilities and build proof-of-concept exploits. The role also involves reverse engineering, investigating 0-day and n-day exploitation, and developing automated tooling to process and analyse data at scale, with results published on the blog or at conferences.

CrowdStrike · Israel

Security researcher role on CrowdStrike Intelligence's Advanced Research Team, performing vulnerability research and exploit development across a wide range of targets. Day-to-day work includes analyzing vendor patches, identifying exploits in large file collections, reverse engineering, and building automated data collection and analysis tooling, with published research on the blog or at conferences.

Anthropic · London · £325,000 – £485,000 pa

Owns the full lifecycle of accelerator capacity, from ingesting and provisioning compute across major clouds and custom datacenters to scaling clusters and building health, diagnostics and repair automation for GPU, TPU and Trainium nodes. Sets technical strategy for node lifecycle management and drives cross-team initiatives to keep one of the industry's largest AI compute fleets reliable and usable.

Wayve · London

Lead a small team iterating on Wayve's end-to-end L2 driving model, balancing hands-on Python coding and experiments with directing priorities. You'll root-cause on-road issues and fix them through data recipes, model architecture, evaluation and release readiness, working across teams on model and vehicle interface problems.

Wayve · London

Develops embedded software connecting an AI Driver to vehicle steering, braking and propulsion systems, working hands-on with benches and vehicles. Involves MATLAB/Simulink development, drive-by-wire actuation logic, CAN/CAN-FD diagnostics, OTA updates and ADAS/HMI interfaces, collaborating with OEMs, suppliers and controls engineers.

Corecom Consulting · Belfast · £25,500 pa

A two-year academy programme for career changers with no prior experience or degree, combining full-time hybrid training with client project work. Trainees cover Agile, SQL, APIs, Python, testing, data science and AI/ML, then work as an AI Engineer and Software Quality Engineer on real client projects.

Luminance · Cambridge

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.

Luminance · Cambridge

Designs and develops the core AI platform, building large-scale Python codebases and web-hosted services that power the company's software architecture. Works across the Python ecosystem, including Flask, Requests, NumPy and PyTorch, with scope to lead technical direction or teams. Open to graduates as well as experienced engineers.

Deerfoot Recruitment Solutions · Didcot · £700 – £900 pd

Architect and build multi-agent AI systems end-to-end, combining frameworks like LangGraph or Haystack with bespoke orchestration, state management and tool-calling logic. Embed LLMs and vision-language models into live reasoning, search and task-execution workflows, deploying across cloud and on-premises environments with monitoring, evaluation and guardrail tooling. Set technical standards for agentic AI within the team, working with a biotech and pharma customer.

Carrington Recruitment Solutions · London · £95,000 – £105,000 pa

Hands-on role designing, building and maintaining production AI solutions across a professional services firm, focused on legal service delivery, knowledge management and operational efficiency. Work spans Copilot Studio, Azure AI Foundry and Azure DevOps, with prompt design, grounding strategies, agentic AI and integration into core business systems, alongside senior stakeholders and knowledge teams.

OpenAI · London

Hands-on engineering partner to international public-sector customers, delivering scoped GenAI and agentic use cases from mission need through architecture, evaluations, production and handoff. Works with government agencies, systems integrators and cloud providers on data residency, sovereign and regulated deployments, building prototypes on model APIs and Codex while feeding insights back to Product and Research.

University of Oxford · Oxford · £39,424 – £47,779 pa

Deliver AI safety research for the ARIADNE consortium, a Wellcome-funded generative AI system conducting clinical interviews to assess psychosis risk. Develop LLM supervisor modules that detect delusional reinforcement and inappropriate responses to suicidal symptoms, evaluating them in controlled testbeds with simulated patients and LLM-as-judge scoring calibrated against clinicians.