AI Evaluation Engineer

AI Evaluation Engineer

Rate / Duration / Location

£700-£750/day Inside IR35 | Duration: 4 months | Clearance: BPSS + SC eligible | Location: London, Bristol or Manchester | Hybrid: 2 days onsite per week.

Overview

We're looking for an experienced AI Evaluation Engineer to join a specialist Public Sector Technology team. This hands-on role focuses on building evaluation frameworks, tooling and harnesses for AI systems, particularly LLM and agentic AI solutions. The role requires ownership, experimentation and rapid delivery.

Key Responsibilities

• Design evaluation frameworks and tooling
• Develop evaluation harnesses
• Evaluate model and agentic AI systems
• Define metrics and methodologies
• Build repeatable evaluation processes
• Investigate AI failures
• Work directly with government teams
• Prototype and test solutions
• Present findings and recommendations
• Contribute to strategic direction.

What We Are Looking For

• Strong software engineering experience
• Experience building and evaluating AI systems
• Understanding of LLMs and agentic AI
• Experience with AI evaluation frameworks
• Ability to code independently
• Strong problem-solving and communication skills
• Comfortable working autonomously.

Technical Experience

AI/LLM Evaluation, RAG Evaluation, Ragas, Agentic AI Evaluation, Evaluation Harnesses, LLM Testing and Benchmarking, Prompt Engineering, Python Development, Model and Agent Integration.

Ideal Background

AI Evaluation Engineer, Applied AI Engineer, AI Engineer, LLM Engineer, Harness Engineer, Prompt Engineer or Software Engineer specialising in AI.

Team

Small Public Sector Technology team of approximately 2 to 6 people within a wider organisation.

What Will Make Someone Stand Out

Experience building evaluation tooling, evaluating LLMs, using Ragas, identifying AI failure modes, combining engineering with strategic thinking, and thriving in fast-moving environments.

Summary

Hands-on AI Evaluation Engineer role supporting government AI initiatives through evaluation, testing, benchmarking and framework development.