The AI Engineer Roadmap (2026)

From zero to building AI systems.

This roadmap is not a quick path. Programs like the Stanford AI Graduate Certificate typically take 1 to 2 years part-time. If you’re self-studying, your timeline will depend on a few things:

• Your starting point: If you already have experience with programming or math, you’ll move faster.

• Your time commitment: A few hours a week will look very different from daily focused work.

• Your depth of learning: You can rush through concepts, or you can actually understand them well enough to build with them.

There’s no single timeline. What matters is not how fast you finish, but whether you can actually use what you learn. This roadmap is designed to get you to that point.

• Goals: Become proficient in Python syntax and libraries.

• Free Classes/Resources:

  • Python Roadmap by Jean

• Recommended Book: Automate the Boring Stuff

• Estimated Time: 4-6 weeks

• Goals: Familiarize yourself with popular libraries.

• Classes/Resources:

  • Numpy Tutorial by Jean

  • Pandas Tutorial

• Estimated Time: 2-4 weeks

• Goals: Learn how to query, filter, and analyze data using SQL. Understand how to work with relational databases, including selecting data, joining tables, aggregating results, and writing efficient queries for real-world use cases.

• Classes/Resources:

  • Introduction to SQL by DataCamp (paid)

  • SQL Cheatsheet by Jean

  • W3 School SQL Tutorial

• Estimated Time: 2–4 weeks

• Goals: Master essential Linux commands and shell scripting.

• Classes/Resources:

  • Linux Academy

  • Introduction to Shell by DataCamp (paid)

• Recommended Written Tutorial: Ubuntu’s The Linux command line for beginners

• Estimated Time: 1-2 weeks

• AI For Everyone

• Google AI Professional Certificate

• Microsoft AI & ML Engineering

Get Coursera Plus for 40% OFF 🎉
CLICK HERE

• Goals: Understand what foundation models are and how they work at a high level. Learn how training, post-training, and scaling impact model behavior so you can choose the right model for different use cases.

• Key Topics:

  • Training data

  • Modeling

  • Post-training

• Class:

  • Foundation Models Tutorial, and Why Not to Fine Tune Them by Stanford

  • MIT 6.S191 (2025): Large Language Models (Google) by MIT

• Written Resources:

  • AI Engineering: Building Applications with Foundation Models by Chip Huyen (paid book)

  • [Gopher] Scaling Language Models: Methods, Analysis & Insights from Training Gopher (DeepMind, 2021): Introduced the modern large-scale scaling paradigm and analyzed how performance improves as models grow.

  • [InstructGPT] Training language models to follow instructions with human feedback (OpenAI, 2022): Showed how human feedback and preference optimization transformed raw LLMs into helpful assistants.

• Goals: Understand how models generate text and how sampling controls output style, including randomness, creativity, and determinism. Learn how parameters like temperature and top-p affect model behavior in real applications.

• Resource:

  • Optimize Your AI Models by Matt Williams.

  • Sampling for Text Generation by Chip Huyen: A clear explanation of how decoding works in practice, covering temperature, nucleus sampling, beam search, test-time compute, and why sampling changes output style.

Continue reading this post for free, courtesy of Jean Lee.