AI工程从零开始

Prerequisites

• You can write code (any language; Python helps).
• You want to understand how AI actually works, not just call APIs.

Getting started

Three ways in. Pick one.

Option A — read. Open any completed lesson on aiengineeringfromscratch.com or expand a phase under Contents. No setup, no cloning.

Option B — clone and run.

git clone https://github.com/rohitg00/ai-engineering-from-scratch.git
cd ai-engineering-from-scratch
python phases/01-math-foundations/01-linear-algebra-intuition/code/vectors.py

**Option C — find your level (recommended).** Skip ahead intelligently. Inside Claude, Cursor, Codex, OpenClaw, Hermes, or any agent with the curriculum skills installed:

/find-your-level

Ten questions. Maps your knowledge to a starting phase, builds a personalized path with hour estimates. After each phase:

```bash /check-understanding 3 # quiz yourself on phase 3 ls phases/03-deep-learning-core/05-loss-functions/outputs/

Phase 0: Setup & Tooling `12 lessons`

Get your environment ready for everything that follows.

<details id="phase-1"> <summary><b>Phase 1 — Math Foundations</b> &nbsp;<code>22 lessons</code>&nbsp; <em>The intuition behind every AI algorithm, through code.</em></summary> <br/>

</details>

<details id="phase-2"> <summary><b>Phase 2 — ML Fundamentals</b> &nbsp;<code>18 lessons</code>&nbsp; <em>Classical ML — still the backbone of most production AI.</em></summary> <br/>

</details>

<details id="phase-3"> <summary><b>Phase 3 — Deep Learning Core</b> &nbsp;<code>13 lessons</code>&nbsp; <em>Neural networks from first principles. No frameworks until you build one.</em></summary> <br/>

</details>

<details id="phase-4"> <summary><b>Phase 4 — Computer Vision</b> &nbsp;<code>28 lessons</code>&nbsp; <em>From pixels to understanding — image, video, 3D, VLMs, and world models.</em></summary> <br/>

</details>

<details id="phase-5"> <summary><b>Phase 5 — NLP: Foundations to Advanced</b> &nbsp;<code>29 lessons</code>&nbsp; <em>Language is the interface to intelligence.</em></summary> <br/>

</details>

<details id="phase-6"> <summary><b>Phase 6 — Speech & Audio</b> &nbsp;<code>17 lessons</code>&nbsp; <em>Hear, understand, speak.</em></summary> <br/>

</details>

<details id="phase-7"> <summary><b>Phase 7 — Transformers Deep Dive</b> &nbsp;<code>14 lessons</code>&nbsp; <em>The architecture that changed everything.</em></summary> <br/>

</details>

<details id="phase-8"> <summary><b>Phase 8 — Generative AI</b> &nbsp;<code>14 lessons</code>&nbsp; <em>Create images, video, audio, 3D, and more.</em></summary> <br/>

</details>

<details id="phase-9"> <summary><b>Phase 9 — Reinforcement Learning</b> &nbsp;<code>12 lessons</code>&nbsp; <em>The foundation of RLHF and game-playing AI.</em></summary> <br/>

</details>

<details id="phase-10"> <summary><b>Phase 10 — LLMs from Scratch</b> &nbsp;<code>22 lessons</code>&nbsp; <em>Build, train, and understand large language models.</em></summary> <br/>

</details>

<details id="phase-11"> <summary><b>Phase 11 — LLM Engineering</b> &nbsp;<code>17 lessons</code>&nbsp; <em>Put LLMs to work in production.</em></summary> <br/>

</details>

<details id="phase-12"> <summary><b>Phase 12 — Multimodal AI</b> &nbsp;<code>25 lessons</code>&nbsp; <em>See, hear, read, and reason across modalities — from ViT patches to computer-use agents.</em></summary> <br/>

</details>

<details id="phase-13"> <summary><b>Phase 13 — Tools & Protocols</b> &nbsp;<code>23 lessons</code>&nbsp; <em>The interfaces between AI and the real world.</em></summary> <br/>

</details>

<details id="phase-14"> <summary><b>Phase 14 — Agent Engineering</b> &nbsp;<code>42 lessons</code>&nbsp; <em>Build agents from first principles — loop, memory, planning, frameworks, benchmarks, production, workbench.</em></summary> <br/>

Each Phase 14 workbench lesson (31-42) ships a mission.md briefing the agent before it opens the full lesson docs.

</details>

<details id="phase-15"> <summary><b>Phase 15 — Autonomous Systems</b> &nbsp;<code>22 lessons</code>&nbsp; <em>Long-horizon agents, self-improvement, and the 2026 safety stack.</em></summary> <br/>

</details>

<details id="phase-16"> <summary><b>Phase 16 — Multi-Agent & Swarms</b> &nbsp;<code>25 lessons</code>&nbsp; <em>Coordination, emergence, and collective intelligence.</em></summary> <br/>

</details>

<details id="phase-17"> <summary><b>Phase 17 — Infrastructure & Production</b> &nbsp;<code>28 lessons</code>&nbsp; <em>Ship AI to the real world.</em></summary> <br/>

FIG_002 · A worked sample

Phase 14, lesson 1: the agent loop. ~120 lines of pure Python, no dependencies.

def run(query, tools):
    history = [user(query)]
    for step in range(MAX_STEPS):
        msg = llm(history)
        if msg.tool_calls:
            for call in msg.tool_calls:
                result = tools[call.name](**call.args)
                history.append(tool_result(call.id, result))
            continue
        return msg.content
    raise StepLimitExceeded