We are seeking a hands-on Senior AI Engineer to deliver a high-profile, time-boxed AI Proof of Concept for a UK regulated payments client. The role focuses on designing and building generative AI agents on Microsoft's stack — Copilot Studio under the client's ratified SDLC Framework, or M365 Copilot Agents via Agent Builder as a fallback.
The engineer will work
directly with client SMEs across Compliance, Audit, Risk and Rules & Standards through weekly validation cycles, and will own the technical build from prompt design through to structured-output delivery
Required Skills & Experience
Generative AI engineering (essential)
* Hands-on production experience building LLM-based applications — not classical ML or analytics. Comfortable with prompt design, structured output generation, retrieval augmented generation (RAG), embeddings, and hallucination mitigation.
* Practical experience with at least one major LLM API or service (Azure OpenAI strongly preferred, OpenAI direct, Anthropic, Google acceptable).
* Ability to debug LLM-based systems systematically: isolating chunking issues, retrieval failures, prompt issues, and model behaviour.
* Understanding of when an agent platform is the right tool versus when to drop into code, and the limits of declarative agent platforms
Microsoft platform fluency (essential)
* Hands-on experience with Microsoft Power Platform: Power Automate flows, Dataverse, and connectors. Experience with AI Builder is a plus.
* Working knowledge of Microsoft Copilot Studio (preferred) or M365 Copilot Agents (declarative agents via Agent Builder). Candidates without direct Copilot Studio experience but with strong adjacent Power Platform / Microsoft GenAI exposure will be considered subject to a short technical assessment.
* Familiarity with SharePoint as a knowledge source for AI agents, Entra ID authentication, and the Microsoft 365 ecosystem more broadly.
* Working knowledge of Azure services relevant to AI delivery — Azure OpenAI, Azure AI Document Intelligence, and Azure Storage — at consumer-of-API level rather than infrastructure-build level