AI Full-Stack Engineering Course

Portfolio-first AI engineering

Learn the stack by shipping evidence.

This course trains you to build AI applications that another engineer can inspect: runnable code, reproducible environments, data notes, model evidence, prompt/RAG/Agent traces, and clear limits.

Start in 30 minutes View the learning path

13chapters

Core content from tools to open-source model delivery

9main chapters

Built for AI application engineering skills

3languages

English, Chinese, and Japanese maintained together

1project thread

Improved continuously across the course

Follow AI Roads

Please keep an eye on airoads.org as the course continues to improve. If enough learners gather here, we will consider creating a study group so everyone can learn, ask questions, and move through the path together.

Choose Your Starting Move

0.1 Run the quick experience

See input, model, and output before learning terms.

0.2 Prepare the minimum environment

Python, Git, and one project folder are enough for week one.

0.3 Read the capability map

Connect tools, data, models, LLMs, RAG, Agents, and delivery.

0.4 Plan your main route

Choose one project thread, set your pace, then enter Chapter 1.

Main Route

Chapters 1-3 Foundations

Developer tools, Python, data analysis, and the habits that make work reproducible.

Chapters 4-6 Understand Models

Practical math, machine learning, deep learning, training loops, and debugging signals.

Chapters 7-9 Build AI Applications

LLMs, RAG, APIs, tool use, Agents, evaluation, and production-facing traces.

Chapters 10-12 Extend the Stack

Computer vision, NLP, multimodal workflows, AIGC, ethics, and delivery strategy.

Chapter 13 Own the Runtime

Open-source LLM compute routes, local serving, evaluation, LoRA decisions, and reproducible runbooks.

What You Are Building Toward

By the end of the main line, your strongest project should explain one AI application end to end. The goal is not to memorize buzzwords. The goal is to prove that you can turn model behavior into a usable product workflow.

• Setup and README commands: another person can run the work.
• Data, prompt, retrieval, and tool traces: the application has observable behavior.
• Metrics, comparison notes, and failure cases: you can judge quality instead of trusting one demo.
• Safety, privacy, cost, and latency notes: you understand product constraints.
• Screenshots or short demos: a reviewer can understand the user experience quickly.

One Rule

Read briefly, run something, keep evidence. At the end of each stage, you should have something another person can inspect: a README command, a saved output, a metric table, a trace, a failure note, or a small demo. This is also how you turn learning into a portfolio story.

Pass Check

You pass this start page when you have chosen one project thread, one immediate chapter to enter, and one evidence artifact you will keep from the first session.

Evidence to Keep

Keep this page’s proof of learning as a first course evidence card:

Target Role

AI application engineer, AI full-stack builder, or AI-enabled product engineer

Project Thread

one assistant, automation, analysis, or multimodal idea to improve across chapters

First Route

quick experience → minimum setup → capability map → Chapter 1

First Evidence

screenshot, saved output, README command, or short observation note

Expected Output

one main-route plan and one concrete portfolio-grade artifact to start