Attention mechanism in detail
A step-by-step account of queries, keys, values, masking, scaling, softmax, and the final weighted sum.
3 min readStructured learning
Notes turn connected AI concepts into calm, durable learning paths. Pick a topic or begin with the recommended foundation.
Recommended first path
A structured path through attention, multi-head composition, positional information, and encoder-decoder architecture.
View the Transformers curriculumA step-by-step account of queries, keys, values, masking, scaling, softmax, and the final weighted sum.
3 min readHow parallel attention projections create complementary views and recombine them into one representation.
3 min readWhere position, masks, residual paths, encoder blocks, and decoder blocks fit in a complete Transformer.
3 min readFollow five tokens through a manually implemented PyTorch encoder, with equations, runnable code, tensor outputs, and a complete dimension audit.
114 min readBrowse the curriculum
A structured path through attention, multi-head composition, positional information, and encoder-decoder architecture.
From text and bytes to subword algorithms, domain tokenizer training, held-out evaluation, and model compatibility.
Design, test, and improve prompts through worked examples, structured outputs, reasoning strategies, and inspectable workflows.
Understand how networked systems communicate, scale, and stay reliable through concrete requests, worked examples, and design trade-offs.
Adapt pretrained Transformers to new tasks through clear terminology, worked metrics, and complete encoder and decoder training workflows.