Suvradeep Das

Suvradeep Das

he/him

Language models & social good

Blog posts

2025

Dissecting the Attention Mechanism and Transformers

less than 1 minute read

Published:

Transformers have revolutionized NLP, powering models like GPT-4o, Claude, and DeepSeek. But what makes them so effective? The answer lies in their attention mechanism, which enables models to focus on relevant information rather than processing everything equally.

Demystifying Tokenization: The First Step in Building LLMs

less than 1 minute read

Published:

Ever wondered how large language models (LLMs) like GPT-4 understand text? Well, they don’t—at least, not like humans do! Instead, they break text down into tokens, the fundamental building blocks of all AI-generated language.