About this role
Salary: £120,000 - 140,000 per year
Requirements: 4+ years of experience building production softwareExperience shipping LLM-powered systems that have been used by real usersHands-on experience across areas such as RAG, prompting and/or agentsStrong production engineering fundamentals, including an understanding of reliability and incident responseGood technical understanding of LLMs, including concepts such as sampling parameters and embeddingsExperience designing meaningful evaluations for AI/LLM systemsComfortable using agentic development tooling such as Claude Code, Codex or similarExperience working and shipping at paceStrong communication skills with the ability to explain technical decisions clearly and concisely Responsibilities: Build, improve and maintain production LLM systemsImprove the cost and latency of LLM-powered features without sacrificing accuracyDesign and implement evaluation frameworks that can reliably measure whether changes genuinely improve the productInvestigate production failures raised by analysts, identify root causes and implement effective fixesWork across areas including RAG, prompting and agentic systemsPartner closely with Product to shape new functionality and take ideas through to productionUse AI-native engineering tools such as Claude Code as part of your day-to-day development workflowMake sound technical decisions when working through ambiguous engineering problems Technologies: AIClaude CodeLLMRAG More:
We are an exciting AI technology business building and improving production LLM systems. This is a hands-on engineering role focused on real-world AI problems where the technology is already being used in production. You will have the opportunity to influence how the product evolves, working closely with Product to improve quality, reduce cost and latency, build better evaluation frameworks and solve real production failures. We are based in London, Moorgate, and offer a hybrid working model with 2 days per week in the office. Our engineering environment is heavily AI-native, with tools such as Claude Code forming a key part of how we build and ship software.
last updated 39 week of 2026