Track 1 · Foundations
AI/ML Foundations
Just enough of how machine learning works to follow everything else on the path without a math degree. No training a model yourself, no heavy equations. Mental models you can actually reason with.
01
02
03
04
05
06
07
What machine learning actually is
The shift from writing rules to showing examples, and what "learning" really means: fitting a function to data. The one idea everything else stands on.
Training vs inference
The two halves of a model's life, why they have wildly different costs, and why "the model expects 16 kHz mono" is a contract you can't break.
Features vs learned representations
Hand-picked inputs vs features the model discovers itself, and why deep learning won by stopping the hand-picking.
Neural networks, visually
Neurons, weights, and layers as a stack of simple transforms. What the network is and isn't doing.
How models learn: loss and gradient descent
A model improves by rolling downhill on a measure of wrongness. The whole training loop in one picture.
Embeddings: meaning as geometry
Turning words, audio, and images into vectors where distance means similarity. The idea behind search, RAG, and most modern AI.
Why GPUs
Why AI runs on graphics chips: matrix multiplication, parallelism, and memory bandwidth as the real bottleneck.