Agentic AI
by DeepLearning.AI
Andrew Ng's four agentic design patterns, built from scratch in plain Python.
Overview
Andrew Ng's Agentic AI is DeepLearning.AI's flagship self-paced course on agentic workflows, running about nine hours and fifty-five minutes across thirty-one video lessons, seven code examples and eight graded assignments organised into five modules: Introduction to Agentic Workflows, the Reflection Design Pattern, Tool Use, Practical Tips for Building Agentic AI, and Patterns for Highly Autonomous Agents. It is taught vendor-neutrally in raw Python rather than around a single framework, so the four design patterns Ng popularised — reflection, tool use, planning and multi-agent collaboration — are built from first principles before any orchestration library appears. Learners build a research and planning agent, a SQL-generation system that critiques and repairs its own queries, a chart-generation workflow, an email assistant and a multi-agent market-research team, finishing with a customer-service pipeline. Public repositories of the course materials show the working stack: OpenAI models called through aisuite, LangChain, Pydantic for structured output, FastAPI and PostgreSQL for the application layer, Tavily, arXiv and Wikipedia as external tools, and Docker for deployment. Module four spends its time on evaluations, error analysis and optimisation rather than new patterns, which is the part most agent tutorials skip; learners report module one starts slowly before the pace picks up. Ng founded DeepLearning.AI and co-founded Coursera and Google Brain. Free enrolment covers the videos and forum; graded assignments and the certificate require a Pro membership at $25-30 per month.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- ~9h 55m, self-paced (5 modules, 31 video lessons, 8 graded assignments)
- Provider
- DeepLearning.AI
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Implement the reflection pattern so an agent critiques and repairs its own output
- ✓Give an agent tools and let it decide which function to call
- ✓Decompose a complex task into sub-tasks using an LLM planner
- ✓Coordinate specialised agents into a market-research or customer-service pipeline
- ✓Build a SQL-generation agent that validates and fixes its own queries
- ✓Run error analysis and evaluations to find which agent step actually fails
- ✓Judge when an agentic workflow beats a single prompt in production
Highlights
- •Taught vendor-neutrally in raw Python, so the patterns outlive any one framework
- •An entire module on evaluations and error analysis, which most agent tutorials skip
- •Nearly ten hours of material with 31 lessons, 7 code examples and 8 graded assignments
- •Taught by Andrew Ng, who named and popularised the four agentic design patterns
- •Course labs are mirrored in public GitHub repos, so you can inspect the code before enrolling
Who It’s For
Best For
- ✓Python developers building their first production agent
- ✓Engineers who have used an agent framework but never seen the loop underneath
- ✓Technical leads deciding where agentic workflows fit in a product
Prerequisites
- •Intermediate Python — you will be reading and writing real code
- •Basic familiarity with LLM APIs and prompting
FAQ
What is Agentic AI?
DeepLearning.AI's flagship course on agentic workflows, taught by Andrew Ng across five modules and roughly ten hours. It teaches reflection, tool use, planning and multi-agent collaboration in vendor-neutral raw Python, then has you build research, SQL, email-assistant and customer-service agents. For Python developers who want to understand the agent loop rather than only call a framework.
Is Agentic AI free?
Agentic AI offers free content, with paid options for certificates or premium features.
What level is Agentic AI for?
Agentic AI is aimed at a intermediate audience. Recommended background: Intermediate Python — you will be reading and writing real code, Basic familiarity with LLM APIs and prompting.
How long does Agentic AI take?
Expect roughly ~9h 55m, self-paced (5 modules, 31 video lessons, 8 graded assignments). Most learners work through it at their own pace.
What will I learn from Agentic AI?
You'll learn: Implement the reflection pattern so an agent critiques and repairs its own output; Give an agent tools and let it decide which function to call; Decompose a complex task into sub-tasks using an LLM planner; Coordinate specialised agents into a market-research or customer-service pipeline; Build a SQL-generation agent that validates and fixes its own queries; Run error analysis and evaluations to find which agent step actually fails; Judge when an agentic workflow beats a single prompt in production.
Topics
Sources
This page was written from 3 sources, 1 on domains other than deeplearning.ai.