I've been neck-deep in AI and LLMs for a while now, and I kept running into the same problem: every learning resource out there for LLMs either tries to teach you all of deep learning from scratch, or throws you into a sea of random “awesome-LLM” repos, hoping you can connect the dots yourself.
So, I wrote up the roadmap I wish existed a year ago: how to actually learn LLMs and build real things, with none of the bloat. It's geared towards folks with a CS (or practical programming) background who want to skip the endless ML prerequisites and get their hands dirty. The approach:
- Concepts first, then phases, then resources/tools
- Each phase has concrete projects (build an autograd engine, write a mini-GPT, fine-tune with LoRA, etc)
- The goal is to get you actually building and shipping, not just watching lectures
I've been neck-deep in AI and LLMs for a while now, and I kept running into the same problem: every learning resource out there for LLMs either tries to teach you all of deep learning from scratch, or throws you into a sea of random “awesome-LLM” repos, hoping you can connect the dots yourself.
So, I wrote up the roadmap I wish existed a year ago: how to actually learn LLMs and build real things, with none of the bloat. It's geared towards folks with a CS (or practical programming) background who want to skip the endless ML prerequisites and get their hands dirty. The approach:
- Concepts first, then phases, then resources/tools - Each phase has concrete projects (build an autograd engine, write a mini-GPT, fine-tune with LoRA, etc) - The goal is to get you actually building and shipping, not just watching lectures
Hope you guys find it useful.
-Ahmad