Artificial Intelligence Engineer

SR2; Greater London; Teletrabalho; GBP 90 000 - 120 000; Tempo integral

Resumo da oferta

SR2 is seeking an AI/ML engineer to build pipelines that use Claude, OpenAI and other models to scan code for vulnerabilities. You will compare models, develop evidence-based routing, and enhance detection rules, enabling a scalable service across departments.

Required are real experience with LLM APIs, strong Python, and an evidence‑driven mindset. You’ll work independently in a remote, multi‑supplier setup with occasional travel to London.

Qualificações

• Active SC clearance (essential)
• Real experience building with LLM APIs — Anthropic, OpenAI, Bedrock, Azure OpenAI or Vertex
• Strong Python, plus agents, tool calling and structured outputs
• An evidence‑first mindset: you prove things work with data
• You’re a self‑starter who’s happy working independently in a multi‑supplier team

Responsabilidades

• Build pipelines that put Claude, OpenAI and other models to work scanning code and spotting vulnerabilities
• Test, compare and evidence which model performs best for each use case
• Build data‑driven model evaluation and routing across providers — the best model for the job, no lock‑in
• Use AI to speed up detection engineering: writing and testing the rules that spot threats
• Help turn a proven pilot into a service other departments can use

Conhecimentos

Python

LLM APIs

Agents & tool calling

Data-driven evaluation

Security tooling knowledge

Ferramentas

Anthropic API

OpenAI API

Bedrock

Azure OpenAI

Gitleaks/Semgrep/CodeQL

Descrição da oferta de emprego

Initially till March 2027, rolling extensions expected

Primarily remote, occasional travel to London

Get hands-on with the latest frontier models and put them to work defending government systems.

You’ll join a central government cyber security team taking an AI code-scanning pilot that’s already delivering results and turning it into a live product. This is a forward‑deployed, build‑it‑you‑self role.

What you’ll do

• Build pipelines that put Claude, OpenAI and other models to work scanning code and spotting vulnerabilities
• Test, compare and evidence which model performs best for each use case
• Build data‑driven model evaluation and routing across providers — the best model for the job, no lock‑in
• Use AI to speed up detection engineering: writing and testing the rules that spot threats
• Help turn a proven pilot into a service other departments can use

What you’ll bring

• Active SC clearance (essential)
• Real experience building with LLM APIs — Anthropic, OpenAI, Bedrock, Azure OpenAI or Vertex
• Strong Python, plus agents, tool calling and structured outputs
• An evidence‑first mindset: you prove things work with data
• You’re a self‑starter who’s happy working independently in a multi‑supplier team

Bonus points: security tooling (Semgrep, Trivy, Gitleaks, CodeQL), detection rules (Sigma, KQL, SPL), GitHub Actions, or model gateways such as LiteLLM.

Good to know: you’ll apply models, not train or fine‑tune them.