Solve hard problems where AI meets cybersecurity.

Open roles

Engineering Full-time London

Build the controls that inspect AI actions before they reach tools, data, and users.

Apply by email →

Build low-latency services that intercept and evaluate agent and tool calls.
Design integrations for model APIs, agent frameworks, MCP, and observability stacks.
Strong Python or TypeScript and experience shipping reliable cloud services.
Hands-on familiarity with LLM agents, tool calling, or model evaluation.

Security research Full-time London

Break agentic systems before attackers do, then turn useful findings into stronger defences.

Apply by email →

Design adversarial tests for injection, jailbreaks, data leakage, and tool misuse.
Automate attack chains and scoring, then convert failures into regression suites.
Security testing or offensive-cyber experience, with strong Python skills.
Hands-on knowledge of frontier models, agents, and tool calling.