DEV Community

Saurabh Pathak
Saurabh Pathak

Posted on

I wrote a 100-topic AI engineer roadmap that runs offline, with no API key

#ai #python #rag #learning

Every time I wanted to learn RAG, agents or MCP properly, I ended up with twelve browser tabs and no order to read them in. So I wrote the order down, and then wrote the lessons.

ai-engineer-roadmap is 100 topics in 7 phases, from "what is RAG?" to production AI. If you know basic Python, you can start today.

Repo: https://github.com/techsaurabh-394/ai-engineer-roadmap

What is in it

Every topic has the same shape:

  • A plain-English explanation with an everyday analogy
  • A diagram
  • Python code
  • 3 interview questions with answers
  • Links to the original paper or official docs

The 7 phases are RAG Foundations (20 topics), Advanced RAG (20), LLM Concepts (14), Agentic AI (16), Frameworks (8), MCP (7) and Production AI (15). Topics are marked beginner, intermediate or advanced, so you can skip what you know.

Try it in 5 minutes 

git clone https://github.com/techsaurabh-394/ai-engineer-roadmap.git
cd ai-engineer-roadmap
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
pytest -q
python 01-rag-foundations/code/mini_rag.py "Why use multi-stage builds?"
Enter fullscreen mode Exit fullscreen mode

The last command runs a complete, tiny RAG pipeline. It chunks three sample documents, scores them against your question and prints the best passage with its source file. The demo scripts in each code/ folder use only the Python standard library, so there is no API key and no network needed. If you set ANTHROPIC_API_KEY and install anthropic, Claude writes the answer from the retrieved passages instead.

What it does not do

  • The demos are small on purpose. mini_rag.py scores text with simple matching, not a real embedding model. It teaches the pipeline, it is not a production retriever.
  • The code snippets inside topic pages use real libraries like sentence-transformers, chromadb and langgraph. Those need installing, and some need an API key. Each one is marked.
  • Phase 7 has Dockerfile and Kubernetes YAML examples, but the demos use small in-process stand-ins, so you do not need Docker or Redis just to read along.

How I kept it honest

CI checks that every local link in the Markdown pages points to a file that exists, validates the SVG diagrams, runs the tests and builds the TypeScript MCP example. I also went back and corrected my own claims when they did not match the repo: an earlier version said 47 interview questions, and the real count is 43 plus 4 system-design prompts.

Who it is for

  • Beginners who keep hearing "RAG", "agents" and "MCP" and want to understand them
  • Developers moving into AI engineering who want a path instead of random tutorials
  • People preparing for interviews. There is a separate interview-prep folder

Help me fix it

If you find a mistake, a broken link, or a topic that is missing, open an issue or a pull request. The glossary and the study guide are good places to start. If it helps you, a star helps other people find it.

Top comments (0)