I self-taught myself RAG & AI Agents

And now I am building RAG & AI Agents in production.

You're wasting time if you're not preparing for production!

..

Here's the roadmap I'd follow today:

1️⃣ 𝗠𝗮𝘀𝘁𝗲𝗿 𝗣𝘆𝘁𝗵𝗼𝗻 𝗳𝗼𝗿 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻

→ Most AI systems run on Python - write clean, maintainable code

→ Learn debugging, logging, testing, and deployment fundamentals

✅ CS50P – Harvard's Intro to Python

🔗 cs50.harvard.edu/python

✅ 100 Days of Code: Python (Udemy)

🔗 udemy.com/course/100-da…

✅ The Missing Semester – Shell, Git, Linux basics

🔗 missing.csail.mit.edu

..

2️⃣ 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 𝗙𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹𝘀

→ Build from scratch, understand backprop and optimization

→ Foundation for debugging any LLM or deep learning system

✅ Neural Networks: Zero to Hero – Karpathy

🔗 karpathy.ai/zero-to-her…

✅ Deep Learning Specialization – deeplearning.ai

🔗 coursera.org/specializa…

..

3️⃣ 𝗚𝗲𝘁 𝗛𝗮𝗻𝗱𝘀-𝗢𝗻 𝘄𝗶𝘁𝗵 𝗟𝗟𝗠𝘀 & 𝗡𝗟𝗣

→ Learn embeddings, tokenization, transformers

→ Essential foundation for RAG and agent systems

✅ Hugging Face Course

🔗 huggingface.co/learn/nl…

✅ Hands-On LLMs – Alammar & Grootendorst

🔗 oreilly.com/library/vie…

✅ LangChain + LLM Projects

🔗 youtube.com/playlist

..

4️⃣ 𝗕𝘂𝗶𝗹𝗱 𝗥𝗔𝗚 + 𝗔𝗴𝗲𝗻𝘁 𝗦𝘆𝘀𝘁𝗲𝗺𝘀

→ Go beyond chatbots - integrate external tools, APIs, memory

→ This is where 80% of production AI applications live today

✅ Introduction to RAG (Coursera)

🔗 coursera.org/learn/retr…

✅ AI Agents with RAG & LangChain (IBM)

🔗 coursera.org/learn/fund…

✅ LangChain for LLM App Development (deeplearning.ai)

🔗 deeplearning.ai/short-c…

✅ LangGraph for Production Workflows

🔗 langchain-ai.github.io/…

..

5️⃣ 𝗦𝗵𝗶𝗽 𝗘𝗻𝗱-𝘁𝗼-𝗘𝗻𝗱 𝗣𝗿𝗼𝗷𝗲𝗰𝘁𝘀

→ Build: PDF Q&A, research assistant, data analysis agent

→ Use production tools: FastAPI, Docker, vector databases

✅ Production RAG Templates & Examples

🔗 github.com/langchain-ai…

..

6️⃣ 𝗟𝗲𝗮𝗿𝗻 𝗠𝗟𝗢𝗽𝘀 & 𝗦𝘆𝘀𝘁𝗲𝗺 𝗗𝗲𝘀𝗶𝗴𝗻

→ CI/CD, monitoring, scaling, failure handling

→ The difference between demos and production systems

✅ Designing ML Systems (Chip Huyen)

🔗 oreilly.com/library/vie…

✅ Made With ML – End-to-end MLOps

🔗 madewithml.com

✅ System Design for ML Engineers

🔗 educative.io/courses/ma…

..

𝗞𝗲𝘆 𝗶𝗻𝘀𝗶𝗴𝗵𝘁: Most people stop at step 4.

Steps 5-6 separate hobbyists from engineers.

The gap isn't knowing transformers.

It's knowing how to deploy, monitor, and scale them.

..

You don't have to do everything!!!!!

Pick what aligns with your goals.

But if you want production-ready skills in 2025, don't skip the engineering fundamentals.

Share this with someone who needs a production-focused AI roadmap 💚