What it does

This repo is a curated syllabus for learning deep learning without drowning in the firehose of MOOCs, books, and YouTube channels. It maps out five phases—prerequisites, fundamentals, building projects, deep dives, and continuous learning—linking only free resources. Think of it as a human filter for the paradox of choice.

The interesting bit

The author doesn’t just dump links. They sequence them deliberately: MIT calculus before 3Blue1Brown for rigor, then 3Blue1Brown for intuition; Andrew Ng’s theory before fast.ai’s code so the latter actually clicks. The “incomplete” framing is honest—it’s a living document, not a finished product.

Key highlights

• All resources are free (explicitly stated, though “as of now” leaves wiggle room)
• Heavy emphasis on prerequisite math with specific concept checklists (multivariable calculus, matrix operations, Bayes theorem)
• Practical GPU workarounds: Google Colab, Kaggle Kernels, AWS student credits
• Explicit target audience split: researchers need more math, practitioners can skate by with less
• Inspired by Haseeb Qureshi’s blockchain guide, borrowing the “authoritative but incomplete” tone

Caveats

• README is truncated mid-sentence; full depth of later phases (4, X) is cut off
• “As of now, all resources are free”—no mechanism mentioned for tracking if links go paid or dead
• Curated in 2018; fast.ai has since revamped courses, PyTorch dominates over Keras—staleness risk is real

Verdict

Worth bookmarking if you’re early in your deep learning journey and paralyzed by options. Skip it if you already know your way around PyTorch and arXiv; this is a map for the lost, not a reference for the arrived.