We are building AI-assisted automation for a large HR and payroll programme: agents and pipelines that turn business requirements and test-case workbooks into test cases, validated staging data and reviewer-ready evidence. The core is already working and in use. We need a hands-on AI engineer to build, host and enhance these agents with us.
* Build agents. Design and implement LLM-based agents and tools: tool and function calling, structured outputs, multi-step workflows, human-in-the-loop approval, prompt and context design.
* Build the tool layer. Extend our MCP (Model Context Protocol) server and APIs that agents call to read requirements, agreements, process maps, reference test cases and product documentation.* Host and operate. Take agents from a local, single-user setup to a hosted service: containerised backend, identity and access control, audit logging, secrets management, deployment on Azure.
* Retrieval and knowledge. Build retrieval over HR and pay documents (collective agreements, process catalogues, test libraries): chunking, embeddings, hybrid search, re-ranking, with measured retrieval quality.* Evaluate. Build labelled evaluation sets and regression tests for every AI feature; report accuracy, precision and recall; keep deterministic logic deterministic where a rule beats a prompt.
* Integrate. Connect outputs to Azure DevOps test management, and build Microsoft 365 agents (Copilot Studio, Agent Builder, Power Automate, Teams, SharePoint) for reviewer feedback and Q&A inside the tenant.* Deliver with the tooling. Run the pipelines on new requirement workbooks, resolve what they flag, reconcile counts, and prepare outputs for review.
* Bring ideas. Propose and prototype new uses of AI across HR and pay testing: coverage analysis, scenario generation, defect triage, process-to-requirement mapping.
* Python, 5+ years. Production code on a shared codebase: typing, pytest, git branches and pull requests, code review.* LLM application engineering, 12+ months shipped. Direct experience with at least one major model API (Anthropic, OpenAI, Azure OpenAI): tool use, structured output, streaming, token and cost management, prompt caching.
* Agentic systems. Built multi-step agents with tools, memory and approval gates; familiar with MCP and at least one agent framework or SDK.* Retrieval (RAG). Embeddings, vector stores, hybrid search, chunking strategy, and how to evaluate retrieval before building on it.
* Evaluation. Labelled datasets, metrics, regression suites for non-deterministic systems; can explain why a number moved.* Backend and hosting. FastAPI or equivalent, Docker, REST, authentication and authorisation, logging, deployment to Azure (App Service, Container Apps or AKS), CI/CD.
* Data handling. Messy multi-sheet Excel and CSV at scale with openpyxl or pandas; reconciliation that adds up; SQL.* AI-assisted development. Works fluently with AI coding assistants day to day and reviews their output critically.
Nice-to-have skills* Azure DevOps: test plans, suites, work items, REST API.
* HCM or payroll platforms (Dayforce, Workday, SAP SuccessFactors, PeopleSoft) or public-sector HR and pay.

Closing: December 05, 2026

Do you have any questions about this vacancy?

Vacancy reference: CR-307599

SB

Corporate recruiter

Steven Berkun-Drevnig

+1 647 288 2762

Contact