Track
build-ai-models
How to Make AI Models
36 lessons: choosing and assembling models—linear models, trees, nets, attention, heads, and latency—plus 108 MCQs.
- Mode
- none
- Practice
- Read / quiz
- Lessons
- 36 units
Before you start
How to choose and assemble a model: linear models, trees, networks, embeddings, attention, and an output head that matches the label.
The wrong family memorizes or cannot represent the decision. A smaller model you can explain often ships.
Feature design, architecture reviews, and the spec you hand to someone who will train the model.
Start from the output type. Compare against a baseline. Write shapes, the head, and a latency budget.
After /data-science/intro—use it with /train-models/intro when you are deciding what to fit, not only how to fit it.
Lesson order
Sequential — follow top to bottom36 lessons are live in this track. Start from step 01 for the smoothest path.
-
01 intro A model is a function with parameters
beginner
Open → -
02 problem-to-model Match the model to the problem
beginner
Open → -
03 linear-models Linear and logistic models
beginner
Open → -
04 trees Decision trees
beginner
Open → -
05 ensembles Ensembles
beginner
Open → -
06 neural-net-anatomy Anatomy of a neural net
beginner
Open → -
07 layers Layers
intermediate
Open → -
08 activations Activation functions
intermediate
Open → -
09 embeddings Embeddings
intermediate
Open → -
10 cnns Convolutional models
intermediate
Open → -
11 sequences Sequence models
intermediate
Open → -
12 attention Attention
intermediate
Open → -
13 transformers Transformers
intermediate
Open → -
14 encoders Encoders
intermediate
Open → -
15 decoders Decoders and generation
advanced
Open → -
16 tokenizers Tokenizers
advanced
Open → -
17 output-heads Output heads
advanced
Open → -
18 classification-vs-generation Classification versus generation
advanced
Open → -
19 regression Regression models
advanced
Open → -
20 ranking Ranking
advanced
Open → -
21 multimodal Models with more than one input type
advanced
Open → -
22 small-models Small models on purpose
advanced
Open → -
23 from-scratch-vs-library Writing it versus using a library
advanced
Open → -
24 shapes Shapes and broadcasting
advanced
Open → -
25 forward-pass The forward pass
advanced
Open → -
26 custom-layer A custom layer
advanced
Open → -
27 loss-design Designing a loss
advanced
Open → -
28 model-cards Model cards
advanced
Open → -
29 size-and-latency Size and latency
advanced
Open → -
30 distillation Distillation
advanced
Open → -
31 quantization-preview Quantization preview
advanced
Open → -
32 architecture-eval Evaluating an architecture choice
advanced
Open → -
33 wrong-model Signs you chose the wrong model
advanced
Open → -
34 documentation Document the model you made
advanced
Open → -
35 interview Interview review
advanced
Open → -
36 capstone Specify a model you could train
advanced
Open →