Your structured path from ML foundations to production agentic AI systems — what to learn, build, and certify.
Core math, Python data stack, ML concepts. You must be able to build and evaluate a basic classifier before moving on.
Understand transformers, use OpenAI/Azure OpenAI APIs, master prompting techniques. Build your first LLM-powered app.
Build end-to-end RAG pipelines. Go beyond naive RAG to hybrid search, reranking, and evaluation with RAGAS.
Build single agents with ReAct, then graduate to CrewAI / LangGraph multi-agent systems. Azure AI Agent Service for production.
Deploy at scale on Kubernetes/AKS, set up CI/CD with GitHub Actions, observability, cost controls, governance.
| Skill Area | Beginner | Practitioner | Senior Engineer |
|---|---|---|---|
| LLM APIs | Can call OpenAI API | Token optimization, function calling, streaming | Fine-tuning, model routing, cost modeling |
| RAG | Naive RAG with Chroma | Hybrid search, reranking, RAGAS eval | GraphRAG, multi-index, production-grade pipelines |
| Agents | LangChain ReAct agent | LangGraph, CrewAI, tool design | Multi-agent systems, custom orchestrators |
| Infrastructure | Docker basic | K8s/AKS, GitHub Actions CI/CD | Terraform, OIDC/WI, multi-region, GitOps |
| Observability | print() debugging | LangSmith tracing, Prometheus metrics | Full observability stack, SLO/SLA, cost attribution |
| Security | API key awareness | Key Vault, managed identity, content filters | Zero-trust, RBAC, prompt injection defense, audit |
AI/ML concepts, Azure AI services overview, responsible AI principles. Entry point to Azure AI.
Implement AI solutions: Azure OpenAI, AI Search, Document Intelligence, Speech, Vision, Bots.
Azure ML, model training, deployment, MLflow, AutoML, Responsible AI at scale.
Tip: Complete path — AI-900 → AI-102 → Build portfolio projects → DP-100 if MLOps focused. GitHub Copilot cert (AZ-140 variant) coming 2025.