Welcome back to Path to Staff! This is the final part of the Unpacking AI series, where we cover Training, Post-Training and Inference, and watch it all come together.
I also crosspost this to my personal Substack where you can subscribe if you are interested in more technical posts like these. After this piece, we will return to regular programming on career growth on Path to Staff.
Here’s the series thus far:
The Hardware Behind AI. Where we talked about hardware, including transistors, fabricators, and most famously the memory-compute bottleneck (why RAM is so expensive!).
Model Architecture. Where we covered the “Attention is All You Need” paper, talked about tokens, embeddings, attention, and covered the transformer block. This is a must-pre-read if you haven’t read it. You’ll need to understand this in order for this piece to make sense.
Training, Post-Training, and Inference. (This essay.) In this piece, we dig into how the numbers inside a model get set, refined and finally return …
