We’re looking for an AI Workflow Engineer / Automation Architect to design and build end-to-end AI-enabled business workflows . You’ll productionize solutions using C#/.NET on Azure Functions , integrating LLMs cleanly and safely into existing systems.
What You’ll Do
- Design and implement AI-powered workflows from idea to production
- Decide what should be handled by LLMs vs deterministic code
- Build reliable, observable, and scalable automation on Azure
- Ensure quality, safety, and continuous improvement of AI workflows
Must-Have Requirements
1. Strong C# / .NET Engineering
- Production-grade C# with modern .NET
- Experience with async programming, dependency injection, error handling, and structured logging
- Writes maintainable code with unit/integration tests
- Familiar with CI/CD and code review best practices
2. Azure Functions & Azure Services
- Hands-on experience with Azure Functions (HTTP, queue, Service Bus triggers)
- Familiar with Durable Functions (if applicable)
- Practical experience with:
- Azure Service Bus
- Azure Storage
- Key Vault
- Application Insights / monitoring
- Understands configuration, secrets, and scaling concerns
3. LLM & Prompt Engineering
- Designs prompts with structured outputs (schemas, JSON)
- Builds guardrails to reduce hallucinations and ambiguity
- Implements validation, retries, fallbacks, and human-in-the-loop where needed
- Understands real-world LLM failure modes
4. AI Workflow Orchestration
- Breaks business processes into clear automation steps
- Implements orchestration patterns such as:
- Idempotency
- Correlation IDs
- Retries and dead-lettering
- End-to-end observability
5. Quality & Evaluation
- Defines evaluation datasets (“golden sets”)
- Runs regression tests for prompts and workflows
- Monitors quality drift and improves performance over time
Nice-to-Have
- Power Platform (Power Automate / Power Apps)
- Document AI / OCR (Azure Document Intelligence or similar)
- RAG / embeddings and retrieval-based AI solutions
- Security & governance experience:
- PII handling
- Prompt injection awareness
- Audit trails and data controls
Originally posted on Himalayas