AlgoCademy
Start learning free

CoursesData & AIMachine Learning

Analyze a shop’s data, then predict who buys again.

For people who know Python and have never used numpy or pandas. You clean and analyze the shop’s data, train models with scikit-learn and test them honestly, write a neural network from scratch, then use PyTorch.

Start the first lesson See the plans
You type every line A real checker runs your code Nothing to install
Lessons
165
Chapters
17
Language
Python
Level
Intermediate
Interview questions
20
Free
The first chapter
Before you start
Python: lists, dictionaries, loops and functions.

What you build

Analysis reports

Best products, revenue per month, best customers and cohorts.

A model that predicts who buys again

With a baseline to beat, cross-validation and no leakage.

A neural network from scratch

Forward pass, loss, gradients and backpropagation, until it learns XOR.

A digit reader in PyTorch

A training loop, batches, dropout and convolutions.

Your own small language model

A tokenizer, self-attention and a transformer that makes up new names.

A forecast and a recommender

Next month’s revenue, and products for one customer.

A /predict API

Your model behind a route that refuses bad input.

Tools and topics

Syllabus

01numpy8 lessons
  1. Your First Arrayfree
  2. Pick Items and Slicesfree
  3. Math on Whole Arraysfree
  4. Filter with True and Falsefree
  5. Rows, Columns and axisfree
  6. Shapes: reshape and Stackingfree
  7. Random Numbers You Can Repeatfree
  8. Project: Put Columns on One Scalefree
02pandas14 lessons
  1. Open a Table: read_csv
  2. Columns and Their Types
  3. Pick Columns and Rows
  4. Filter Rows
  5. Sort and Take the Top
  6. Make New Columns
  7. Missing Values
  8. Clean Messy Text
  9. Count What's in a Column
  10. Group and Summarize
  11. Several Summaries at Once
  12. Join Two Tables: merge
  13. Dates
  14. A Table of Two Things: pivot_table
03Analysis Projects8 lessons
  1. Which Products Sell?
  2. Revenue per Month
  3. Who Are the Best Customers?
  4. Who Comes Back?
  5. Cohorts: Customers by Join Month
  6. Buckets: Age Groups
  7. Outliers
  8. What Moves Together?
04Your First Model10 lessons
  1. Features and a Label
  2. Hold Some Rows Back
  3. Fit, Then Predict
  4. How Often Is It Right?
  5. Beat the Baseline First
  6. Four Kinds of Right and Wrong
  7. Precision and Recall
  8. Not Yes or No: A Probability
  9. Predict for a New Customer
  10. Project: Your First Model, End to End
05Models14 lessons
  1. Predict a Number: Linear Regression
  2. Read What the Model Learned
  3. How Wrong, on Average?
  4. Logistic Regression
  5. Put Features on One Scale
  6. Decision Trees
  7. How Deep Should a Tree Grow?
  8. Random Forests
  9. Which Features Matter?
  10. Nearest Neighbors
  11. Words into Numbers: One-Hot
  12. Groups Without a Label: k-means
  13. How Many Groups?
  14. Project: Pick the Best Model
06Doing It Right10 lessons
  1. Overfitting
  2. Cross-Validation
  3. Keep the Mix: stratify
  4. Leakage: Answers Hiding in the Features
  5. Fit Only on Training Rows
  6. Pipelines
  7. Different Columns, Different Steps
  8. Tune the Settings
  9. When One Answer Is Rare
  10. Save and Load a Model
07A Neural Network from Scratch9 lessons
  1. One Neuron
  2. Squeeze It: Sigmoid
  3. A Layer Is a Matrix
  4. The Forward Pass
  5. How Wrong? The Loss
  6. Which Way Is Down? A Gradient
  7. Learn: Gradient Descent
  8. Backpropagation
  9. Project: A Network Learns XOR
08ML Interview Questions20 lessons
  1. Accuracy, Precision, Recall from Scratch
  2. Write train_test_split
  3. Write a Standard Scaler
  4. Linear Regression with Gradient Descent
  5. Logistic Regression from Scratch
  6. k-Nearest Neighbors from Scratch
  7. k-means from Scratch
  8. Build a Confusion Matrix
  9. Cosine Similarity
  10. Softmax
  11. Gini Impurity
  12. Find the Best Split
  13. Naive Bayes from Scratch
  14. AUC Without a Library
  15. Write k-Fold Cross-Validation
  16. PCA from Scratch
  17. Linear Regression in One Line of Algebra
  18. TF-IDF from Scratch
  19. Write a One-Hot Encoder
  20. Choose the Threshold
09More Models7 lessons
  1. Gradient Boosting
  2. How Many Trees, How Big a Step
  3. Stop When It Stops Helping
  4. The Fast One: HistGradientBoosting
  5. What Really Matters: Permutation Importance
  6. Fewer Columns: PCA
  7. PCA Inside a Pipeline
10Text7 lessons
  1. Words into Counts
  2. Common Words Count Less: TF-IDF
  3. Is This Review Happy?
  4. The Words That Decide
  5. Two Words Together
  6. Naive Bayes
  7. Find Similar Texts
11Time Series7 lessons
  1. Revenue per Week
  2. Smooth It: Rolling Averages
  3. Yesterday as a Feature: Lags
  4. Split by Time, Never at Random
  5. Two Baselines to Beat
  6. A Model on Lags
  7. Project: Forecast Next Month
12Recommenders6 lessons
  1. The Baseline: Most Popular
  2. A Customer by Product Table
  3. Bought Together
  4. Similar Products
  5. Recommend for One Customer
  6. Is It Any Good? Hold One Out
13Statistics and A/B Tests8 lessons
  1. Center and Spread
  2. A Sample Is Not the Truth
  3. How Sure? The Bootstrap
  4. An A/B Test: Two Conversion Rates
  5. Could It Be Luck? A Permutation Test
  6. What a p-value Does Not Say
  7. How Many Visitors Do You Need?
  8. Don't Peek
14Serving a Model6 lessons
  1. A /predict Route
  2. Send the Probability Too
  3. Refuse Bad Input
  4. Many Customers in One Request
  5. Train Once, Load at Start
  6. Project: Know What You Predicted
15Deep Learning with PyTorch12 lessons
  1. Tensors
  2. Gradients for Free: autograd
  3. Train a Line by Hand
  4. nn.Linear and nn.Module
  5. A Loss and an Optimizer
  6. The Training Loop
  7. Batches: Dataset and DataLoader
  8. Read Handwritten Digits
  9. Test It: eval and no_grad
  10. Overfitting and Dropout
  11. A Network That Looks: Convolutions
  12. Save, Load, and Where a GPU Fits
16Build Your Own LLM14 lessons
  1. Text into Numbers
  2. How Tokenizers Learn: One Merge
  3. Guess the Next Letter by Counting
  4. Inputs and Targets
  5. Embeddings
  6. The Smallest Model That Learns
  7. Look Back, Never Ahead
  8. Self-Attention
  9. Several Heads
  10. A Transformer Block
  11. Assemble the Model
  12. Train It
  13. Generate: Temperature
  14. Project: A Name Generator, and What the Real Ones Add
17Capstone5 lessons
  1. Capstone: Look Before You Model
  2. Capstone: Build the Features
  3. Capstone: Baseline and a Model
  4. Capstone: Test It Honestly
  5. Capstone: Say What You Found

Questions

Something else?
team@algocademy.com

Do I need a GPU?

No. The models are small and train in seconds in the checker. The last PyTorch lesson shows where a GPU fits.

More in Data & AIAI EngineeringData EngineeringSQL & DatabasesAll courses

LESSON 1

Your First Array

The first chapter is free, no card needed. The rest comes with PRO.

Start the lesson See the plans