Architect and build the agentic orchestration layer on top of Microsoft 365, designed to be LLM-agnostic from the start
Set up and configure Azure infrastructure (resource groups, Entra ID app registration, Azure AI Search, storage)
Integrate the Claude API with a locked system prompt; implement interaction logging so all 25 users contribute to a shared organisational knowledge base
Connect Word and Outlook via Microsoft Graph in Phase 1; expand to PowerPoint, Teams, SharePoint, and Excel in Phase 3
Build branded document output templates (logo, colour palette, formatting) that lock in on all AI-generated exports
Implement role-based permissions, search guardrails, and an admin control panel
Provide regular project check-ins with the co-founder (1–2 hour daily overlap window)
Advise on scope and timeline — if something needs more time, say so early

3+ years of experience building on Microsoft Azure — resource provisioning, Entra ID / Azure AD, app registrations, security policies
Hands-on experience with the Microsoft Graph API for M365 integrations (Word, Outlook, SharePoint, Teams)
Experience integrating LLM APIs (Claude, OpenAI, or similar) into production systems, including prompt design and governance
Familiarity with RAG pipelines and vector/AI search (Azure AI Search or equivalent)
Ability to design model-agnostic orchestration architectures that can swap underlying LLMs without rebuilding the stack
Experience with interaction logging, knowledge base construction, and indexing pipelines
Independence: comfortable owning a greenfield project with limited internal technical support on the client side
Communication: fluent English, proactive with updates, clear when scoping risks arise
Accountability: you set the expectations, you deliver on them

Remote contract role — work from anywhere with a reliable connection
Flexible hours with a 1–2 hour daily overlap window for project check-ins (no fixed timezone required)
Hourly rate, with scope to extend beyond the initial 3-month engagement if the project grows
Direct access to the co-founder — no layers of management, fast decisions
A genuinely interesting technical problem: greenfield AI infrastructure on M365 for a firm with real institutional data to work with

AI Solutions Engineer, Architect

Key skills