Build AI From Scratch

This path takes you from AI curiosity to actually building language models. Not just using APIs - understanding what’s happening inside and implementing it yourself.

Prerequisites: Some programming experience recommended. Complete “Become a Developer” first if you’re starting from zero.

Phase 1: Prepare Your Mind

AI involves complex concepts. Make sure you can learn efficiently and think clearly about abstract problems.

Learning How to Learn

Memory techniques, focus strategies, and study optimization. Essential for tackling dense technical material.

Thinking Clearly

AI hype is full of fuzzy thinking. Build the mental tools to cut through noise and understand what’s actually happening.

Phase 2: Build a Language Model

The main event. Go from zero ML experience to a working language model you trained yourself.

Building an LLM From Scratch

• Quick start - working model in 30 minutes
• Transformer architecture explained
• Modern techniques (RoPE, RMSNorm, SwiGLU)
• Training at scale
• Deployment for real use

Time investment: ~8 hours to go from zero to deployed model.

What You’ll Have

After completing this path:

• A working language model you built and trained
• Deep understanding of transformer architecture
• Ability to read and understand AI research papers
• Foundation for further AI/ML specialization

Beyond This Path

This path focuses on understanding. For production AI work, you’ll want to explore:

• Fine-tuning and RLHF
• Prompt engineering and RAG
• ML infrastructure and MLOps
• Specialized architectures (vision, audio, multimodal)

But understanding comes first. Once you know how these systems actually work, learning the rest becomes much easier.