01
AI · Explained08 episodes · free
AI engineering, from zero to hero.
A season-based video course for developers. We start by building a working language AI in pure Python — then climb, one short animated lesson at a time, to embeddings, backpropagation, and transformers. New episodes every Monday, Wednesday, and Friday.
Watch the full playlist on YouTube →S1
Build a Tiny Language AI
Start here. We build a working sentence-completion AI in pure Python and NumPy — no frameworks — and watch it answer something we never taught it.
02
Inside the Network
4:1803
How It Learns
4:4904
The Surprise
4:59S2
How AI Represents Data
Before a model can think, text must become numbers. Embeddings, similarity, and tokenization.
05
What Are Vector Embeddings?
2:0806
How AI Finds Similar Things
2:0707
Tokenization: How Text Becomes Numbers
2:38S3
How Networks Learn
The training loop from the inside: neurons, layers, gradient descent, and backpropagation.
08
What a Neuron Actually Computes
2:2509
Why Deep Networks Are Deep
coming soon
10
Gradient Descent, Visualized
coming soon
11
Backpropagation: the Chain Rule, Made Visual
coming soon
S4
Word Embeddings from Scratch
The hands-on sequel: words stop being isolated slots and start having relationships.
12
Beyond One-Hot
coming soon
13
Learning the Embedding Table
coming soon
14
What the Space Learned
coming soon
15
Using Embeddings for Real
coming soon
S5
Modern AI
Attention, transformers, LLM training, RAG — how today's systems actually work.
16
Attention: How AI Decides What Matters
coming soon
17
The Transformer, Assembled
coming soon
18
How LLMs Are Actually Trained
coming soon
19
RAG: Embeddings + LLMs, Working Together
coming soon
20
Why LLMs Hallucinate
coming soon
Want this kind of clarity applied to your business? Book a free AI audit — I'll show you where AI actually pays off in your workflow.