ML & Vision

Machine Learning Playbook & Deep Learning Lab

Interactive machine learning educational playground and hands-on laboratory covering computer vision, regression, and deep neural nets.

Read the deep dive Source

Overview

The Machine Learning Playbook is an extensive educational curriculum and hands-on software suite covering classical machine learning, statistical modeling, and deep neural networks. Containing interactive web apps and complete Jupyter implementations, it takes learners from regression and classification through to convolutional neural networks (CNNs) for MNIST digit classification and model serving.

Design decisions

The choices that shaped the system, and why.

1
Interactive Canvas Digit Recognition
Created a live drawing canvas in the web interface so users can test trained MNIST models with real-world mouse/touch inputs.
2
Model Quantization for Edge Serving
Converted heavyweight model checkpoints into quantized ONNX formats for instant browser inference.

Architecture

Multi-Part Curriculum

Spans supervised learning, clustering algorithms, natural language processing basics, and deep learning architectures.

Interactive Web Application

Lightweight interactive UI allowing users to draw digits and run real-time inference against trained CNN models.

Production Model Serving

Reference code for packaging PyTorch/Keras weights into REST APIs for production deployment.

Data Visualization Pipeline

Matplotlib and Seaborn scripts for confusion matrices, ROC-AUC curves, and decision boundary plots.

DevOps & deployment

Executable on Google Cloud Vertex AI, AWS SageMaker Studio, or local GPU workstations.

Features

Hands-on MNIST handwritten digit classification lab with high accuracy model.
Comprehensive tutorials on feature scaling, cross-validation, and hyperparameter tuning.
Pre-trained weights and interactive web demo for live model inference.
Clean, reproducible notebooks with complete mathematical explanations.

Screenshots

Related writing

Deep dive · 7 min

ML 63: DL Project 2 MNIST Digit classification using NN

The "hello world" of deep learning — recognizing handwritten digits — because it's simple enough to train fast, yet real enough to teach every core neural-network concept.

22 May 2026

Deep dive · 7 min

ML 64: DL Project 3 Dog vs Cat Classification Transfer Learning

The final lesson of the course, and it teaches the single most practical deep-learning skill: reusing a model someone else already trained on millions of images, instead of starting from zero.

22 May 2026

Deep dive · 5 min

ML 60: Diabetes model for API

A Streamlit app is great for a human clicking through a browser — but other software (a mobile app, another server) needs an API instead. This is that shift in thinking.

21 May 2026

Deep dive · 9 min

ML 61: DL Project 1 Breast Cancer Classification with NN

Your first neural network — the same classification problem you already solved with Logistic Regression, now solved by a model that can learn much more complex patterns.