ML, AI, and Agentic Development: From Foundations to Tools

A course about how machine learning, deep learning, and modern AI actually work — and how to build agentic systems on that foundation. Aimed at a developer who wants not just to “call an API” but to understand the engine under the hood: how models learn, what neural networks and transformers are, where large language models came from, and how to assemble an agent from a model that acts in the world through tools.

The course deliberately complements two others in the collection. The Data Science course provides applied classical ML (pandas, scikit-learn, evaluation) — here we go deeper into the principles: optimization, neural networks, deep learning. The LLM course teaches you to build applications on top of ready-made models (prompting, RAG, API) — here we explain why those models work and rise to the engineering of agents as a discipline. Together, the three courses give the full picture: apply, understand, build.

Two parts. Foundation (lessons 1–7): the map of AI/ML/DL, how models learn (optimization), learning paradigms, neural networks, deep learning architectures, transformers and LLMs, foundation models and generative AI. Agents and engineering (lessons 8–12): what an agent is, tools and the MCP protocol, patterns and frameworks for agentic systems, evaluation and safety, production. The tooling landscape is given as of mid-2026 (MCP, LangGraph, agent orchestration) — but the principles matter more than the specific libraries, which change monthly.

Each lesson contains theory, examples (diagrams, pseudocode, Python), a checklist, and a quiz. The through-line: modern AI isn’t magic but an understandable engineering tower of ideas, each of which can be mastered. Understanding the foundation distinguishes someone who deliberately builds reliable systems from someone who copies other people’s code and hopes it works.