Generative AI Engineer
Agentic AI Engineer (.NET) — Contract
- Client: Global Technology Leader – NASDAQ Listed
- Location: Hybrid — South West UK / 1-2 Days Per Week
- Duration: 6 months + Intention to extend (Greenfield Multi-Year Programme)
- Day rate: £500 — Inside IR35
- Start date: ASAP
Our client, a global leader in software systems, is building AI agents that generate production .NET code under developer supervision. The goal of this greenfield is a spec-driven pipeline where a specification and technical plan drive what the agent builds, and a developer reviews and validates the output at every stage before it ships.
We are looking for a contractor who has built and shipped AI agents in production and who brings a strong .NET background. This is a hands-on engineering role with scope to set the technical direction and bring the wider team up to speed on the approach.
What you will do:
- Design and build agents that produce production-grade .NET code.
- Implement a spec-driven development workflow (spec, plan, tasks, implement), using Spec Kit.
- Build review and approval gates so developers can verify and correct agent output before it is merged.
- Integrate tools and data sources into the agents via MCP (Model Context Protocol).
- Set up evaluation, guardrails and tests to catch incorrect output early.
- Work alongside engineering team’s and upskill them on agentic patterns and the spec-driven approach.
Essential experience:
- Proven track record building and shipping AI agents in production, including tool calling, orchestration and multi-step workflows. Prototype-only experience will not be enough.
- Strong .NET and C# background (.NET 8/9, ASP.NET Core).
- Hands-on with a .NET-native agent framework, such as Microsoft Agent Framework or Semantic Kernel.
- Experience building agents whose output is code, with a clear focus on validating and verifying correctness.
- Familiarity with spec-driven development, ideally Spec Kit.
- Experience integrating tools via MCP.
- Designing human-in-the-loop review and approval gates within agent workflows.