Principal Ai Architect, Azure & Agentic Systems - Canada (Full Remote) - Up To 200 000$ + Bonus

Principal AI Architect, Azure & Agentic Systems Remote, Canada (anywhere in Canada, home office, occasional travel to clients and partner events) · Permanent, full-time employee · Principal level, 8+ years · up to 200 000 CAD base salary plus performance bonus
InCycle Software is hiring a Principal AI Architect in Canada to own the architecture behind a new generation of Azure-native AI and agentic systems for our enterprise clients. You will define how large language models, agent frameworks and Azure services come together into production platforms that hold up under real traffic, real data and real security reviews. This is a hands‑on principal role: you set the technical direction, write the reference implementations the engineering teams build on, and make the calls that keep these systems reliable, secure, observable and affordable.
About the CompanyAbout InCycle InCycle Software is a Microsoft Solutions Partner for Digital & App Innovation and Data & AI, and a GitHub Verified Partner. We work with enterprise and mid-market clients across Canada and the United States on four things:
cloud strategy and governance,
application modernization,
modern engineering and DevOps, and
data, analytics and AI, including agentic AI strategy and GenAI solution accelerators built on Azure AI Foundry.
Our consultants are senior engineers first. They work in small teams on client engagements, often alongside Microsoft, and they are trusted to make technical decisions and see them through to production. We believe there is always a better way to build and deliver software, and we are hired to find it.
ResponsibilitiesYou are the senior technical authority on AI and agentic work across InCycle's portfolio. On client engagements, you lead the architecture from the first strategy session through the reference implementation and the handover, for platforms built on Azure AI Foundry, Azure OpenAI and the surrounding Azure services, deployed in Canadian regions when data residency requires it. Internally, you own the AI reference architectures, accelerators and standards the rest of the team delivers with, and you decide which parts of a fast‑moving ecosystem InCycle adopts and which it waits out. A typical engagement runs three to nine months and might be an agentic customer‑service platform for a financial services client, a RAG system over a manufacturer's engineering documents, a public‑sector assistant that must stay inside Canadian borders, or an evaluation and guardrail layer added to an AI product a client already shipped and now cannot trust. You will usually be on one or two engagements at a time, with clients in Canada and, on occasion, the United States. You are the person the CTO sends into the room when a customer asks whether their agent architecture will survive contact with production. You report to the Practice Lead, Data & AI.
What you will do
Design end-to-end architecture for Azure-hosted AI platforms, including single-agent, multiagent and human-in-the-loop orchestration patterns, and decide when an agent is the right answer and when a plain workflow is.
Lead solution design on Azure AI Foundry, Azure OpenAI and the Foundry Agent Service, with retrieval‑augmented generation (RAG), Azure AI Search and vector indexing as first‑class components rather than bolt‑ons.
Set the standards for prompt and tool orchestration, function calling, Model Context Protocol (MCP) integrations, agent memory, evaluation and guardrails, and make them real through templates and reference implementations.
Architect the services around the models: AKS and Azure Container Apps, Azure Functions, API Management, Event Grid and Service Bus, Cosmos DB and Azure SQL, with identity on Entra ID and secrets in Key Vault.
Design for Canadian data residency and sovereignty where clients require it, using Canada Central and Canada East, model availability by region, private endpoints and the governance controls that keep data and inference in country.
Build the operating layer that most AI projects skip: observability for LLM calls and agent traces, offline and online evaluation, cost and latency budgets, content safety and responsible‑AI governance that would pass an enterprise security review.
Own the cost, latency and reliability trade‑offs of LLM inference and vector workloads at scale, including model selection, caching, batching, provisioned throughput and fallbacks.
Partner with product, data and delivery leads to turn business goals into technical roadmaps that engineers can execute in weeks, not quarters.
Mentor senior and staff engineers through design reviews, pairing and proofs of concept, and raise the bar for how the whole team builds with AI, including how it uses GitHub Copilot and agentic development tools itself.
Represent the technical vision with customers and partners, including Microsoft Canada, in architecture reviews, briefings and the occasional conference talk.
Qualifications
Bachelor's degree in Computer Science or equivalent experience.
Legal authorization to work in Canada.
Required Skills
What you bring - 8+ years in software engineering and architecture, with recent, hands‑on delivery on Azure;
you still write code and can show it.
Proven experience architecting LLM or agentic systems that are in production today, serving real users, with the scars to prove it: hallucination handling, tool‑call failures, runaway costs, prompt injection, evaluation drift.
Deep command of the Azure platform: compute, networking, identity, data and the AI services, including Azure AI Foundry, Azure OpenAI, Azure AI Search and the content safety and monitoring tooling around them.
Working knowledge of the current agent ecosystem: Semantic Kernel, AutoGen or the Microsoft Agent Framework, LangGraph or similar, and MCP, with the judgment to know which to adopt and which to skip.
Strong command of at least one of C#/.NET, Python or TypeScript, with modern API design, containers, Kubernetes (AKS), CI/CD on GitHub or Azure DevOps, and infrastructure as code in Bicep or Terraform.
Solid grasp of RAG, embeddings, chunking and retrieval quality, evaluation methods, and AI safety and guardrail practices, and the ability to explain them to a CISO as well as to an engineer.
The presence to influence technical and executive audiences in writing and in the room, in clear English, and the professionalism to represent InCycle in front of client leadership. French is an asset for engagements in Quebec and with federal clients.
#J-18808-Ljbffr