Agentic AI Full-Stack Senior Engineer

Senior Agentic AI Engineer (Full-Stack) | Azure | .NET Angular

I'm partnering with a fast-growing SaaS business to find a Senior Agentic AI Engineer who builds production systems, not demos. If you have shipped multi-agent workflows into real customer environments and care deeply about latency, token cost, reliability, and security, this role was written for you.

What you'll be doing

• You'll design and ship agentic products end to end, from orchestration logic through to the operator UI.
• That means building multi-agent workflows using AutoGen and Semantic Kernel, grounding them with Azure AI Search and Azure OpenAI, and owning the full delivery cycle from R&D through to stable production. On the back end you'll be writing .NET Core/C# services and robust tool APIs.
• On the front end, Angular and TypeScript for operator dashboards with human-in-the-loop controls and audit visibility.
• Security isn't a checklist item here. PII-safe prompt handling, output validation, data minimisation, and safe function calling are implemented as code and tests from day one.

What we're looking for

• Hands-on production experience with AutoGen or Semantic Kernel (both strongly preferred), including tools/functions implementation and measurable outcomes across latency, cost, and reliability.
• Practical experience with Azure OpenAI and Azure AI Search in production retrieval and function-calling patterns. A track record of instrumented, observable, maintainable systems, not impressive-looking prototypes.
• Preferred on top of that: CrewAI, LangGraph/LangChain, MCP server build and consumption, experiment tracking and A/B evaluation for LLM features, and familiarity with frontier models beyond OpenAI (Claude, Gemini, Llama, Mistral).

The stack

AutoGen, Semantic Kernel, Azure OpenAI, Azure AI Search, .NET Core/C#, Angular/TypeScript, Azure API Management, SQL Server, Cosmos DB, AKS, Azure DevOps, GitHub Actions, Application Insights, Prometheus/Grafana.

What good looks like in the first 90 to 180 days

An agentic workflow in production within 8 to 12 weeks, with agreed p50/p95 latency and per-task token cost budgets met. Increasing tool-use success rates and evaluation pass rates. Zero PII leakage in tests, with guardrails documented and enforced in code.

The setup

Hybrid role. Full-time on-site for the first six months as you get fully embedded with the team, then flexibility from there.

Interested or know someone who might be? Drop me a message or reach out directly at g.streefland@franklinfitch.com.