Starting the LLM Journey
Beginning to dive into how large language models actually work.
I’ve started diving into LLM training — really understanding what’s happening under the hood, not just calling APIs.
Where I began
Andrej Karpathy’s LLM101n course is my starting point. His ability to explain complex concepts simply is exactly the kind of teaching I aspire to.
Key insight so far
The transformer architecture is elegant. Self-attention isn’t just a mechanism — it’s a fundamentally different way for a model to “understand” relationships in data. Every token gets to “look at” every other token and decide what’s relevant.
What’s next
- Finish the course
- Implement a small transformer from scratch
- Fine-tune a model on my own data
- Document everything here
More updates soon.