A practical, code-first curriculum for becoming a Generative AI Engineer; from transformers and embeddings through LangChain, RAG, and LangGraph agents, in 42 articles across 7 phases, each paired with a short companion video.

• What this is
• Who it's for
• How this series works
• Curriculum
  • Phase 1: Foundations of Generative AI
  • Phase 2: Working with LLMs
  • Phase 3: Vector Databases & Embeddings
  • Phase 4: LangChain
  • Phase 5: Retrieval Augmented Generation (RAG)
  • Phase 6: LangGraph & Agentic AI
  • Phase 7: Deployment & Capstone Projects
  • Bonus capstone projects
• Repository layout
• Where to go next
• Notes

42 articles, organized into 7 phases, taking a reader from "what is generative AI" to shipping a deployed, agentic RAG application. Every phase builds on the one before it for example, embeddings before vector databases, vector databases before RAG, chains before agents, so the series is meant to be read in order rather than sampled topic by topic.

Each article follows the same shape: the concept explained in plain language first, then a full, runnable implementation. Nothing is left as "an exercise for the reader"; every idea introduced gets code that actually produces it.

New data science and machine learning learners moving into generative AI engineering. It assumes the basic knowledge of Python, Machine Learning, Model training but no prior exposure to LLMs, LangChain, or vector databases.

• Python 3.9+
• A free-tier account with an LLM provider (OpenAI and/or Anthropic) from Phase 2 onward
• A Google Colab account (or a local GPU) for the hands-on fine-tuning article in Phase 2
• No prior LangChain, vector database, or agent experience required

Every numbered article in the tables below ships as:

• a written article (theory plus a complete, working implementation), and
• a video covering the same material. Each article also carries a format tag:

Bridging machine learning into generative AI — the concepts every later phase depends on.

Moving from theory to building — calling, running, evaluating, and adapting real LLMs.

The storage and retrieval layer every RAG system in Phase 5 is built on.

The orchestration framework that connects every component built so far.

The single most important pattern in production generative AI.

Stateful, multi-step agents that plan, reason, and self-correct.

Shipping real applications and building a portfolio that proves the skill.

Three additional, self-contained projects — not counted in the 42 core articles — for extra portfolio pieces once Phase 7 is complete.

• Phase 1 — Foundations of Generative AI (0/6)
• Phase 2 — Working with LLMs (0/7)
• Phase 3 — Vector Databases & Embeddings (0/5)
• Phase 4 — LangChain (0/8)
• Phase 5 — Retrieval Augmented Generation (RAG) (0/7)
• Phase 6 — LangGraph & Agentic AI (0/6)
• Phase 7 — Deployment & Capstone Projects (0/3, + 0/3 bonus projects)

A suggested layout — adjust folder names to match however the articles end up hosted (this repo directly, GitHub Pages, or an external blog):

.
├── phase-1-foundations/
│   ├── 01-what-is-generative-ai.html
│   ├── 02-transformer-architecture.html
│   └── ...
├── phase-2-working-with-llms/
├── phase-3-vector-databases/
├── phase-4-langchain/
├── phase-5-rag/
├── phase-6-langgraph/
├── phase-7-deployment-capstone/
├── bonus-projects/
│   ├── document-qa-system/
│   ├── ai-coding-assistant/
│   └── research-agent/
└── README.md

Each article folder is meant to hold the article itself alongside any notebook, script, or dataset it references, so a reader can clone the repo and run the code for that article without hunting for dependencies elsewhere.

This series assumes a prior Machine Learning series as its foundation and ends with a deployed, agentic RAG application. If you know basic fundamentals of Machine Learning and Deep Learning, natural next steps are:

1. This series. Generative AI fundamentals through LangChain, RAG, and LangGraph.
2. The bonus capstone projects. Three additional builds to round out a portfolio.
3. Specialized production topics. Evaluation at scale, multi-agent orchestration, and MLOps for LLM-powered systems.

• Articles are added to the tables above as they are published.
• The series is meant to be followed in phase order: later phases assume the vocabulary and code from earlier ones (Phase 5's RAG pipeline reuses the vector stores from Phase 3 and the chains from Phase 4, for example).
• Phase 2 onward requires API access to at least one LLM provider (OpenAI and/or Anthropic); Phase 2's fine-tuning article additionally needs a GPU, available for free through Google Colab.