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

Session	Name	Topic	Status
F1	The Prediction Engine	Regression & Supervised Learning	✅ Ready
F2	Cluster Thinking	K-Means & Unsupervised Learning	🔜 Coming soon
F3	The GenAI Builder	GenAI Foundations	🔜 Coming soon
F4	Ship Ready	EDA, Business Thinking & Sprint Readiness	🔜 Coming soon

---

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 In ChatGPT: Paste the SKILL.md contents and say "follow these rules for our session"

---

Research

research/regression-supervised-learning.md — the full conceptual foundation behind F1. The 13 thinking frameworks and 8 AI coding agent moments that the F1 skills encode.

---

The idea behind this repo

AI coding assistants are good at writing code. They are bad at judgment.

They default to binary classifiers when you need ranking. They use random splits on time-series data. They report RMSE to stakeholders who need rupee impact. They add polynomial features instead of asking a domain expert.

These skills don't make the AI write more code. They make it ask better questions. The SKILL.md files encode the judgment of a senior ML engineer — the questions they ask before writing a line, the checks they run after training, the translations they make before presenting results.

One session. One skill. Better engineering.