Practical understanding of transformer architectures and how modern LLMs work.
Hands-on experience with MCP (Model Context Protocol).
Agentic orchestration using frameworks such as Semantic Kernel, LangChain/LangGraph, Microsoft Copilot Studio, OpenAI/Anthropic tool-calling patterns, or equivalent.
Prompt engineering: systematic prompt design, few-shot learning, chain-of-thought, and prompt optimization for foundation models.
Vector databases (e.g., Pinecone, FAISS, Weaviate, Chroma) and RAG architecture patterns.
LLM evaluation methods, including LLM-as-a-judge, automated benchmarking, and regression testing of model outputs.
Post-deployment model maintenance: observability, monitoring, cost optimization, and drift detection.
Developing and implementing governance frameworks and controls for responsible AI.

Experimentation and measurement:

A/B testing, KPI instrumentation, and operational analytics.
Strong Python proficiency; working knowledge of Java or another enterprise language.
Object-oriented design techniques and principles.
Cloud platforms (Azure, AWS, or GCP) and cloud-native development patterns.
CI/CD pipelines and DevOps tools: Git, GitHub, Jenkins, ADO Pipelines, or equivalent.
RESTful API design and integration patterns (JSON, microservices).
Agile methodologies such as Scrum.
Microsoft Office tools (Project, Excel, Word, PowerPoint).
Familiarity with AI-assisted development tools (GitHub Copilot, Cursor, Claude Code) is a plus