1 Introduction to AI (Part 1) A brief history of AI Traditional AI Generative AI Hands-on Activity: Setting up development environment 2 Introduction to AI (Part 2) AI Technological challenges Ethics, bias and regulation in AI AI case studies: social media, health, and national security Discussion: AI use cases for Tanzania government agencies 3 Big Data at the core of AI Big Data at scale: Volume, Velocity, Variety Data structures and optimised operating systems: Relational databases, NoSQL, Hadoop Big Data workflow from collection to final data preparation 4 Deploying AI in institutions Strategic vision from the top Technology and know-how to deploy AI at scale Employee’s skill Set to reap AI rewards Culture of innovation and change inside institutions Discussion: What are some of the challenges Tanzania is facing with AI and Big Data today? How can AI and Big Data be incorporated into the Telecom and IT industry in Tanzania to enhance supervisory efforts? 5 AI Regulation around the world The Europe examples The US example The China examples Tanzania’s approach Discussion: Safeguarding sensitive citizen data in Tanzania 6 Big Data Analysis (Part 1) Query / store Big Data with SQL & Google Big Query Clean and Manipulate data frames with NumPy Visualise Big Data with Jupyter Notebook, Matplotlib, Seaborn & Plotly Hands-On Activity: Learn to create arrays, apply operations, filter and reshape data, and handle missing values using NumPy. 7 Big Data Analysis (Part 2) Big Data distribution: Histograms and kernels Statistical models: Linear regression Interpreting statistical results: Tests and scores Visualise Big Data with Jupyter Notebook, Matplotlib, Seaborn & Plotly Hands-on Activity: Creating visualisations using Matplotlib 8 Big Data Project (Part 1) Collecting real-time data from the web Pre-processing both structured and unstructured data in Database Final data preparation with pandas 9 Big Data Project (Part 2) Project Presentation: Jupyter notebook interactive slides Final debriefs about data complexity and data scalability challenges 10 Machine Learning: Supervised Learning Introduction to Supervised Learning Linear Regression and Logistic Regression Decision Trees Random Forests Hands-on activity: Apply logistic regression to a labelled dataset 11 Machine Learning: Unsupervised Learning Introduction to unsupervised Learning Dimensional Reduction: Principal Component Analysis (PCA) Clustering techniques: K-means, SVM Hands-On Activity: Use Scikit-Learn library to do simple classification tasks 12 Machine Learning: Deep Learning Introduction to Neural networks Why use Neural networks? Neural Networks architecture Introduction to Tensorflow and Keras Hands-On Activity: Build an image recognition model using TensorFlow/Keras 13 Natural Language Processing (NLP) Introduction to NLP NLP applications and use cases NLP architecture: from CNN and RNN to Transformers Transfer Learning in generalised models Models of Transformers used in latest generative AI applications Hands-On Activity: Using an open-source LLM from AWS Sagemaker 14 Machine Learning Project Workflow Collect and Clean data Set-up computing environment: Cloud (Google Cloud/Azure/AWS), VM (Docker) and tooling (MLflow) Work on data pre-processing pipelines Choose model architecture Train model in the cloud Test model with test dataset Deploy model 15 Machine Learning Project Workflow (Part 2) Monitor model performance real-time Set up logging for tracking model behaviour and errors Establish alerts and thresholds for model performance degradation Continually train and update model weights to adapt to new data and requirements Discussion: Propose a project workflow for an educational use case in Tanzania 16 Machine Learning: Deployment at Scale, Security & Operations (MLSecOps) Best practices to deploy AI Projects at scale How to continually retrain models at scale Introduction to MLSecOps architecture Demonstrations of popular MLSecOps tools and tips for tool selection and integration Common attack vectors on ML models and techniques for detecting and mitigating adversarial attacks Real-world case studies highlighting successful MLSecOps implementation for understanding of best procedures. Hands-On Activity: Stress testing of machine learning models through adversarial attacks 17-19 Capstone Project The participants will Implement a ML workflow project from end to end: From collecting raw Big Data to the final UI powered by an AI model deployed at scale Project objective: Create an AI Model for sentiment analysis of Social Media data Stage 1: Collect and clean Social Media data Stage 2: Work on social media data pre-processing pipelines and choose a model architecture Stage 3: Train model to determine the emotional tone of the text (positive, negative, neutral), extract entities (people, places, organisations) and keywords. Stage 4: Create visualizations to represent insights. Use Matplotlib to generate graphs and charts and visualise sentiment distribution, trending topics, word clouds Stage 5: Model evaluation and refinement Stage 6: Presenting models and interpreting insights derived from the models 20 Future Trends in AI and Big Data The new Generative AI wave Open-source vs close to close models Proprietary Big Data Conclusion & Graduation Ceremony