01
Phase 01
LLM Fundamentals
Understand how large language models generate responses, use context, and fit into an application.
120 hour roadmap
Agentic AI engineering roadmap.
A focused 120 hour course for software engineers preparing to build AI applications. Start with LLM fundamentals, then move through retrieval, tools, agentic workflows, and evaluation before completing the Customer Service Agent capstone.
15 hours
02
Phase 02
Prompt Engineering, Context Engineering, and Structured Outputs
Design model requests, assemble trusted context, and return validated data for application code.
03
Phase 03
Retrieval-Augmented Generation (RAG)
Retrieve relevant, current information before an LLM writes an answer.
04
Phase 04
Tool Calling and Model Context Protocol (MCP)
Connect an LLM to tools while application code keeps control of permissions and execution.
05
Phase 05
Agentic Workflows
Design agent workflows with state, controlled decisions, and human oversight where needed.
06
Phase 06
Evaluation, Observability, and Safety
Evaluate AI behavior, trace requests, protect users and data, and manage cost and releases.
07
Phase 07
Customer Service Agent Capstone
Bring the roadmap together in a small public application for customer refund requests.
Plan for around 15 hours each week to complete the course in two months. The estimate includes reading, implementation, debugging, evaluation, and capstone work.