AI Developer: From Zero to Hero in Machine Learning and Deep Learning

Part I: Machine Learning and Deep Learning Fundamentals

Module 1: understanding the theoretical foundations of machine learning

• Key concepts and the history of artificial intelligence and machine learning.
• Different types of machine learning and their challenges.
• Mathematical foundations essential to ML: probability, statistics and linear algebra.
• The complete lifecycle of a machine learning project.

Module 2: understanding the theoretical foundations of deep learning

• Fundamental concepts of deep learning and neural networks (NNs).
• Common neural network architectures: Multi-Layer Perceptrons (MLPs), Convolutional Neural Networks (CNNs) for computer vision, and Recurrent Neural Networks (RNNs) for sequence and text processing.
• Key DL algorithms, including gradient descent.
• Deep learning challenges and limitations.

Part II: Building AI Solutions

Module 3: setting up the AI development environment

• The Python ecosystem for machine learning and deep learning, including SciPy, NumPy and Matplotlib.
• Setting up Python development environments: Virtualenv, Conda, Poetry, Jupyter Notebook and Docker.
• The importance of GPUs for deep learning and how to access them, including free offerings such as Google Colaboratory and Kaggle.
• TensorFlow Keras fundamentals and implementing a first neural network.
• Neural network debugging techniques.

Practical exercises:

• Install and configure the Python ecosystem.
• Train an MNIST classifier on Colab and Kaggle.
• Debug a simple neural network.

Module 4: applying shallow machine learning with Scikit-Learn

• Loading and understanding data for machine learning.
• Data preparation: rescaling, standardisation, normalisation and feature engineering.
• Evaluating ML algorithms and selecting a baseline.
• Using Scikit-Learn pipelines to structure projects.
• Improving model performance, including tuning methods such as Grid Search and Random Search.
• Finalising models and persisting them to disk.

Practical exercises:

• Complete an end-to-end machine learning project.

Module 5: developing deep learning with Scikit-Learn and Keras

• Integrating Scikit-Learn and Keras for deep learning.
• Preparing data specifically for MLPs.
• Defining and compiling MLP architectures: input/output layers, hidden layers, activation functions, loss functions and optimisers.
• Training and evaluating MLPs, including automatic or manual validation and Scikit-Learn k-fold cross-validation.
• Techniques for optimising MLP performance.

Practical exercises:

• Complete an end-to-end deep learning project with an MLP.

Module 6: using Keras professionally

• Model checkpointing and strategies for resuming training after failure.
• Monitoring model training through history and visualising learning curves: Accuracy & Loss.
• Regularisation techniques to address overfitting: Early Stopping and Dropout.

Practical exercises:

• Complete a project incorporating advanced Keras techniques.

Part III: Execution and Deployment

Module 7: packaging and deploying a model as an API

• FastAPI framework fundamentals for building web applications and APIs.
• Serialising and deserialising machine learning models.
• Developing a minimal web application to expose an ML model.
• Packaging the application with Docker.
• Local deployment and deployment to free cloud platforms such as Hugging Face Spaces.
• MLOps fundamentals and limitations of the simple deployment approach.

Practical exercises:

• Build and deploy an MNIST web classifier.

Part IV: Beyond Technology

Module 8: understanding AI for business

• Defining AI in a business context and its main risks: inaccuracy and intellectual property infringement.
• Different approaches to integrating AI into organisations—Takers, Shapers and Makers—and their associated costs (TCO).
• The hidden side of generative AI and intellectual property disputes.
• The history of and opportunities in Silicon Valley.

Interactive discussion (bonus):

• Questions and answers with industry expert Alexandre Z. NICHE, CEO of Datogon. Topics include Datogon's history, starting a business in the United States, life in Silicon Valley, salaries and opportunities in AI.