One day · 6 hours · Hands-on Python

See tensors. Think in ML.

A hands-on session on reshaping, solving and factorizing tensors — written in NumPy, run on real images, video and city data.

2 October 2026 · 5:00 PM–8:00 PM COT (UTC-5) · Bogotá

Read the handbook →

Set up in Colab

No install — every notebook runs in Google Colab.

Pre-course · Start here

Prerequisites study deck

Find your gaps before the day, not during it: the linear algebra the workshop builds on, free, in English and Spanish.

  • Podcast
  • Quiz game
  • Live quiz
  • Flashcards
  • Practice test
  • and more

Open the study deck All preparation steps

Taught in English · Questions in English & Spanish

Open the widget: three photos as a 4-D tensor, transposed and then reshaped

Real photos, one cube per byte. Transpose keeps the picture; reshape breaks it.

One number → a vector → a matrix → a tensor → a layer. Click a picture to try it live.

From one number to a neural-network operation A scalar becomes a vector, then a matrix, then a three-dimensional tensor, each labelled with its NumPy shape. The tensor feeds a neural-network layer that turns inputs into outputs. THE SHAPE OF MACHINE LEARNING 7 ScalarVectorMatrix ()(4,)(3, 3) 3D tensor (3, 4, 4) Neural network y = σ(Wx + b) inputoutput

By the end, you’ll be able to…

  • Own the shape.

    Slice, reshape, and broadcast tensors in NumPy.

  • See the structure.

    Decompose tensors. Deconvolve a real photo in a live deep dive.

  • Connect it to ML.

    Read the tensor math behind compression and neural networks.

Workshop byRavi Kalia and Sebastian Laverde Chunza

Workshop repositoryExplore the code on GitHub ↗ Live discussionSmall groups on Discord ↗

Before the workshop

Vectors and matrices are what you need to begin. Complete the required steps before the session.

Step 1: Core Concepts Required · ~2–3 h

Review vectors and matrices in Goodfellow Ch. 2 and complete the pre-work notebooks. That chapter’s section on SVD is worth a skim if it is new to you — the workshop re-introduces it from scratch, so it is not required reading. Then check yourself with the prerequisites study deck: a podcast, a quiz game, a live quiz, flashcards and a practice test on the same chapter.

Step 2: Setup Required, 5 mins

Run Notebook 00 in Google Colab to check that Colab and the data downloads work on your machine. Colab blocked, or no Google account? See the FAQ.

Step 3: Visual Refresher Optional · ~3 h

Watch 3Blue1Brown’s Essence of linear algebra.

On the day: bring your laptop and charger, sign in to Google in your browser, and open Notebook 00 from the notebooks page when the session starts.

Who this is for

Software engineers entering ML

Practice reading tensor shapes in NumPy and recognize the same operations in PyTorch and JAX.

Data analysts and scientists

Extend familiar tabular workflows to images, sequences, and higher-dimensional arrays.

Self-taught ML practitioners

Trace shape mismatches through indexing, reshaping, broadcasting, and contractions.

STEM graduates

Connect linear algebra to working Python examples of decompositions and ML.

Your workshop toolkit

📖 Handbook

Full theory, exercises, and step-by-step solutions.

Read Handbook

🚀 Notebooks

Interactive Python notebooks running in Google Colab.

Run Code

📊 Slides

Visual presentation decks for workshop sessions.

View Decks

💡 Companion Hub

Mind maps, 1-minute shorts, audio overviews, and self-checks.

Explore Hub

🎯 Kahoot

Live interactive quizzes for concept checks.

Join Quiz

📚 References

Core papers, textbook chapters, and repository source code.

View Sources

🗂️ Study decks

Flashcards, practice tests and quiz games at your own pace, before or after the session: one deck for every section, and one for the prerequisites. Free to start, in English and Spanish.

Start Studying Prerequisites Deck

THE LEARNING JOURNEY

Your path from arrays to ML.

Six connected ideas. Code them. See them. Make them click.

  1. 01

    Foundations

    Read shapes and axes.

  2. 02

    Tensor Operations

    Reshape. Broadcast. Contract.

  3. 03

    Linear Algebra

    Solve systems. Find structure.

  4. 04

    Convolutions

    Turn pixels into features.

    Live deep dive
  5. 05

    Factorizations

    Compress without losing the plot.

  6. 06

    ML Applications

    Connect the math to models.

Explore the notebooks

Try one in the browser

Explore the complete set of widgets, from broadcasting and image layouts to attention, audio, and factorization. No installation or account needed.

See them all

6 hours. One hands-on rhythm.

  1. 10m Code
  2. 5m Concept Review
  3. 10m Group Discussion

The day is thirteen sections, three Kahoot checks and four deep dives opened live — attention, convolution, CP and AlphaTensor — with short breaks and a lunch the clock does not count. The handbook’s schedule has every start time.