DataCamp 实地评测 — Tech With Tim 跑完 20+ 小时
别看数百个 AI Engineer Roadmap 了——跟着一条跑通就行。Tim 亲自花 20+ 小时跑完 DataCamp "Associate AI Engineer for Developers" 课程，得出：这是目前最扎实、最系统的入门路径之一（$13/月），但不完美（前期偏慢 / LLM Ops 浅 / 缺综合 capstone）。
"AI engineering is a lot less about training ML models or doing statistics and mathematics and much more about applying things that have already been built."
OpenAI / Anthropic / 开源 LLM
检索增强生成
从 demo 到 production
开源模型 + transformer pipeline
框架 + agent 编排
重在实现，不是训练
入门起薪 $150K+/年，是 2026 年增长最快的开发岗位之一
| # | 课程名 | 核心内容 | Tim 评价 |
|---|---|---|---|
| ① | Working with OpenAI API | 请求 / model selection / response API / system/user/assistant messages | 基础 |
| ② | Prompt Engineering ⭐ | one-shot / multi-shot / few examples / structured outputs | Tim 收获最大 |
| ③ | Project 1: OpenAI 多轮对话 | 存储 context / token tracking / cost 计算 | 实战 |
| ④ | Working with Hugging Face | 开源模型 / text classification / image classification / auto classes / tokenization / Q&A pipeline / PDF extraction | 系统 |
| ⑤ | LLM Ops | 高层概念（开发 vs 运维）/ 开源 vs 闭源 / fine-tune 决策 | 偏浅 |
| ⑥ | Developing AI Systems with OpenAI | HF + OpenAI 组合应用 | 难度上升 |
| ⑦ | Embeddings + Vector DB + RAG ⭐ | embedding models / Chroma DB / Pinecone / RAG 实战 | 核心 |
| ⑧ | LangChain + AI Agents | 链式调用 / tool calling / 复合 agent | 实战 |
| ⑨ | 高级轨道（选修） | LLM concepts / 高级主题 | 延伸 |
概念讲解，1.5-2x 速度看完
test cases 必须 pass 才能进
不会就回退查
主动学习 > 被动观看
"I don't feel this covers everything to become an AI engineer. But it gets you halfway there, which is a good starting point."
这门课覆盖的内容正好是 hybrid router 的实操层：
学完 = 能用 LiteLLM 搭生产级 hybrid workflow
"AI engineering is applying things already built. You don't need math. You use pre-trained models to create applications that deliver real value."
"This is one of the best resources I found. I have personally tested and tried it. It is very high quality and goes through content in a way that's very digestible and interactive to learn."
从 0:00 到 15:14 的关键节点