AI Engineer

numi; Manchester; Hybrid; GBP 60,000 - 90,000; Full time

Job summary

numi is building a brand-new AI function from the ground up in Manchester, with a genuine greenfield scope and no legacy debt. This is a founding role in a global consumer business, working hybridly three days in the office to shape architecture and delivery from day one.

You will design, build and operate AI agents and automations across Copilot Studio, Foundry and Claude, owning end-to-end delivery, testing and handover, while collaborating with business stakeholders and AI analysts.

Qualifications

• 2–6 years in applied AI, software or automation engineering.
• Active interest in AI with hands-on projects.
• Architecture-level judgement for design decisions.
• Hands-on Copilot Studio and Power Platform.
• Solid Python development practices (testing, version control, CI/CD) and Azure exposure.
• Experience deploying production-grade AI agents with prompts, RAG and evaluation.
• Strong written communication and problem solving.

Responsibilities

• Identify high-value AI and automation opportunities and rapidly prototype POCs.
• Choose platform calls (Copilot Studio, Foundry, Claude) for each opportunity.
• Take proven POCs to production-grade agents with testing.
• Build Copilot Studio and Foundry agents using Python and Azure Functions.
• Develop Power Automate flows with orchestration and error handling.
• Document agents and decisions to a handover-ready standard.
• Collaborate with business stakeholders across regions and suppliers.

Skills

Applied AI

Python

Independent problem solving

Written communication

Tools

Copilot Studio

Azure AI Foundry

Claude

Power Automate

Dataverse modelling

Azure Functions

CI/CD

Job description

Build a brand-new AI function from the ground up inside a global, operationally complex consumer business. Genuine greenfield — no legacy estate, no technical debt.

Location: Manchester — hybrid, 3 days in office

Team: Founding member of a new AI function

Experience: 2-6 years applied AI / automation

The opportunity

You’ll work directly with the business and AI Business Analysts to turn real opportunities into working proofs of concept. Where a POC proves out and drives measurable value, it scales into production — and the team scales with it, with growth pencilled in for 2027. The architecture decisions you make in the first few months are what the function is built on. This is a foundations role, not a maintenance one.

The role

Not a traditional software engineering role. You’ll design, build and operate practical AI agents, automations and orchestration across Copilot Studio, Azure AI Foundry and Claude. This is a sole‑builder role: you own delivery end‑to‑end — architecture, platform choice, build, testing and handover — working across functions, shops, markets and suppliers. You’ll be mentored technically by the Automation Manager but work independently as the hands‑on architect for the AI agent estate.

What you’ll do

• Work with the business to find high-value AI and automation opportunities, and build working POCs fast to test feasibility and value.
• Make and defend the platform call (Copilot Studio vs. Foundry vs. Claude) for each opportunity.
• Take proven POCs to production‑grade agents, including all required testing and validation.
• Build Copilot Studio agents (low‑code) and pro‑code Foundry agents in Python, Azure Functions and custom workflows.
• Develop Power Automate flows with real orchestration — error handling, retries, long‑running operations.
• Own the architecture of the AI agent estate: retrieval strategy (RAG / live query), eval sets, guardrails, observability and CI/CD.
• Document every agent and promotion decision to a handover‑ready standard, building cost‑effectively to maximise ROI.

What you’ll bring

• 2-6 years in applied AI, software or automation engineering, with a portfolio or real production examples.
• A genuine, active interest in AI — already building agents, workflows or tools in your own time.
• Architecture‑level judgement: able to make and defend design decisions without a peer group to lean on.
• Hands‑on Copilot Studio and wider Power Platform (Dataverse modelling, security roles).
• Python with sound engineering habits (testing, version control, CI/CD) and hands‑on Azure, ideally Function Apps.
• Practical LLMs in production — prompt orchestration, RAG and output evaluation.
• Excellent written communication and structured fault‑finding.

Nice to have: Azure AI Foundry (Agent Service, Prompt Flow), Azure AI Search, Semantic Kernel / Microsoft Agent Framework, an infrastructure/cloud background, ML/forecasting, and certifications AB-620 or AI-102/103.