Course

How LLMs Actually Work

0/4 modules complete0%

Modules

4 total

Trace raw data through pretraining and post-training to a deployed assistant checkpoint.

Explain why one API endpoint now handles classification, summarisation, and generation.

Explain Transformers as parallel sequence processors during prefill.

I can explain how quantization shrinks an LLM's memory and cost by storing weights at lower precision, and what quality it trades away.