AI and ML Foundations with Python

AI and ML Foundations with Python

Learn to design, build, and apply machine learning and generative AI models through hands-on, scenario-based labs. This course blends core machine learning foundations—data wrangling, visualization, regression, and classification—with modern generative AI techniques using open-source LLMs and Transformers.

Students explore Python’s data-science ecosystem with Pandas, NumPy, and Matplotlib, build and evaluate models with scikit-learn, examine interpretability and ethical AI, and progress into deep learning using PyTorch. By course end, participants will have built and deployed an AI-powered application capable of running locally or in the cloud.

Audience

• Python Developers
• Data Analysts and Aspiring Data Scientists
• DevSecOps Engineers
• AI-Curious Managers and Architects

Prerequisites

• Python – PCEP Certification or equivalent experience
• Familiarity with Linux command-line environment

Learning Outcomes

• Prepare and visualize data with Pandas, NumPy, and Matplotlib
• Train and evaluate classical ML models with scikit-learn
• Interpret models using SHAP and understand responsible AI concepts
• Train and optimize Transformer models using PyTorch
• Master advanced prompt engineering
• Understand Transformer architecture, embeddings, and self-attention
• Work with open-source LLM frameworks such as LLaMA and cpp
• Apply model quantization and fine-tuning techniques
• Compare CPU vs GPU hardware acceleration
• Develop a real-world AI application powered by open-source models

AI and ML Foundations with Python Outline

Foundations of Machine Learning and Data Handling

Python for Data Science Refresher

• Lecture: Python’s role in Data Science and AI
• Lecture: Overview of Pandas, NumPy, and Matplotlib
• Lab: Solve basic data-science problems using Pandas and NumPy
• Lab: Create visualizations (histograms, scatter plots, correlations) with Matplotlib and Seaborn

Exploratory Data Analysis (EDA) and Preprocessing

• Lecture: Understanding datasets, features, labels, outliers
• Lecture: Data wrangling and preprocessing (missing data, normalization, encoding)
• Lab: Clean and preprocess a real dataset (e.g., Titanic or Housing)
• Lab: Visualize distributions and relationships to identify predictive features
• Classical Machine Learning with scikit-learn
• Lecture: Introduction to supervised learning, regression and classification
• Lecture: Unsupervised learning overview, clustering basics
• Lab: Implement linear regression and evaluate R² and MSE
• Lab: Train a classification model (e.g., Logistic Regression or RandomForest) and visualize confusion matrices
• Lecture: Model evaluation metrics, accuracy, precision, recall, F1-score
• Lab: Experiment with k-means clustering and compare cluster performance

Explainability and Ethical AI

• Lecture: Why explainable AI matters, bias, transparency, and trust
• Lab: Use SHAP or feature importance visualization on a small model
• Lecture: Responsible AI practices and data ethics in machine learning

Transition from Classical ML to Generative AI

Deep Learning and the Rise of Transformers

• Lecture: From neural networks to Transformers, key differences
• Lecture: Feed-Forward Neural Networks and the role of embeddings
• Lecture: What is Intelligence? Introduction to Generative AI
• Lecture: The Transformer model explained
• Lab: Tokenization, breaking text into tokens
• Lab: Word embeddings, numerical representation of language
• Lab: Positional encoding, enabling sequence understanding

Build a Transformer Model from Scratch

• Lecture: Introduction to PyTorch
• Lab: Construct a Tensor from a dataset
• Lab: Orchestrate Tensors in blocks and batches
• Lab: Initialize PyTorch generator function
• Lab: Train the Transformer model with positional encoding and self-attention
• Lab: Attach feed-forward neural network and build decoder block
• Lab: Review the complete Transformer model as runnable code

Model Evaluation and Comparison

• Lecture: Comparing classical ML vs deep learning, trade-offs in complexity and data needs
• Lab: Re-run earlier regression/classification datasets using a simple neural network to compare results

Generative AI, Prompt Engineering, and Real-World Applications

Prompt Engineering

• Lecture: Introduction to prompt engineering concepts
• Lab: Getting started with Gemini or OpenAI API (depending on environment)
• Lab: Develop basic prompts and explore role/context patterns
• Lab: Intermediate prompts — define tasks, inputs, outputs, constraints, and style
• Lab: Advanced prompts — chaining, role setting, feedback loops, and exemplars

Hardware Requirements and Optimization

• Lecture: GPUs vs CPUs — why acceleration matters
• Lecture: TensorCore architectures vs legacy GPU designs
• Lecture: Cost and performance analysis for training and inference workloads

Working with Pre-Trained LLMs

• Lecture: A history of neural network architectures
• Lecture: Introduction to the cpp interface
• Lecture: Preparing an A100 GPU for server operations
• Lab: Operate LLaMA2 models with LLaMA.cpp
• Lab: Select quantization levels balancing performance and perplexity
• Lecture: Running the cpp package
• Lab: Experiment with interactive and persistent contexts
• Lab: Constrain outputs with grammars
• Lab: Deploy a LLaMA API server and interact via a client

Fine-Tuning and Application Integration

• Lab: Fine-tune a model using PyTorch on a small dataset
• Lab: Apply advanced prompt-engineering techniques
• Lab: Build a simple web application (Flask or FastAPI) that integrates your fine-tuned or pre-trained model
• Lab: Implement basic error handling and response caching

Testing and Pushing Limits

• Lab: Optimize model parameters and prompt patterns for best results
• Lab: Evaluate latency and resource utilization across different quantization levels

Capstone Project

Build a Real-World AI Application

Students design and deploy a small but functional AI-driven application using open-source LLMs. The capstone incorporates:

• Dataset preparation and EDA from Day 1
• Model fine-tuning and deployment from Day 3
• Integration of ethical AI principles and explainability visualizations

Certification

• Alta3 AI & ML Foundations with Python Certification