AI工程师的技能. 一会一会.

每次会话都是一个文件夹. 每个文件夹包含SKILL.md文件——把它们放入你的项目,让你的AI读取,它开始像高级工程师而不是代码生成器一样表现.

– – . . .

□会话

++会话 名称 ++话题 ++状态 ++ |.

• F1* ** 预测引擎 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ################################################################################################################################################################################################################################################################ GenAI基金会 即将来临
• F4* ** 船已就绪 * EDA、业务思考和冲刺准备 * * 即将到来 *

– – . . .

^ F1 — 预测引擎QQ

9个技能。 完全回归管道。 从"我们建什么"到"这里是商业影响".

– – . . .

^ F2 — 集群思考 ^

** K- Means & 无监督学习—— 当你没有标签时**

计划的技能:无监督框架、K- Means直觉、集群故障模式、 审定(elbow + silhouette),段对视,全无监督管线.

– – . . .

^ F3 — GenAI 构建器 QQ

** GenAI基金会——法学硕士、催化学、RAG、API工程**

AI Engineering OS

Skills for AI engineers. One session at a time.

Each session is a folder. Each folder contains SKILL.md files — drop them into your project, tell your AI to read them, and it starts behaving like a senior engineer instead of a code generator.

---

Sessions

---

F1 — The Prediction Engine ✅

9 skills. Full regression pipeline. From “what are we building?” to “here’s the business impact.”

f1-the-prediction-engine/
├── SESSION.md                     ← Start here
├── problem-framing-skill/         ← Frame the right problem before touching data
├── model-hypothesis-skill/        ← Choose a hypothesis that matches your situation
├── loss-design-skill/             ← Design loss around business costs, not defaults
├── training-diagnostics-skill/    ← Debug training before trusting the output
├── feature-engineering-skill/     ← Domain knowledge beats polynomial noise
├── generalization-skill/          ← Overfitting, regularization, assumption checks
├── data-integrity-skill/          ← Catch leakage before it reaches production
├── evaluation-skill/              ← Report rupee impact, not just RMSE
└── ml-system-skill/               ← The full 7-stage pipeline, gate by gate

---

F2 — Cluster Thinking 🔜

K-Means & Unsupervised Learning — when you don’t have labels

f2-cluster-thinking/
└── SESSION.md                     ← Preview: what's coming

Skills planned: unsupervised framing, K-Means intuition, cluster failure modes, validation (elbow + silhouette), segment-to-insight, full unsupervised pipeline.

---

F3 — The GenAI Builder 🔜

GenAI Foundations — LLMs, prompting, RAG, API engineering

f3-genai-builder/
└── SESSION.md                     ← Preview: what's coming

Skills planned: LLM intuition, prompt architecture, RAG design, API engineering, GenAI vs traditional ML decision framework, building a real tool.

---

F4 — Ship Ready 🔜

EDA, Business Thinking & Sprint Readiness — from model to delivered project

f4-ship-ready/
└── SESSION.md                     ← Preview: what's coming

Skills planned: hypothesis-driven EDA, data quality assessment, insight-to-action, production code habits, Git workflow, sprint tools.

---

How to use any skill

1. Find the skill that matches where you are in your project
2. Copy that SKILL.md into your working folder
3. Tell your AI: Read SKILL.md and follow it
4. The AI asks the right questions instead of jumping to code

In Cursor: @SKILL.md — read this and follow it In Claude Code: Read SKILL.md and follow it for this project