Data Science AI & ML with Gen AI

Learning Objective: Build rigorous core programming skills in Python alongside foundational mathematics and statistics required for data science.

Core Python Programming

• Anaconda

• Variables and operators

• Control flow

• Functions

• Modules and packages

• Type hinting

• Virtual environments

• Logging

Data Structures & Programming Techniques

• Strings, lists, tuples and dictionaries

• Sets and frozensets

• Comprehensions

• Regular expressions (Regex)

• Exceptions

• Iterators and generators

• Decorators

• File I/O

• Exercises and case studies

Object-Oriented Programming

• Classes and objects

• Abstraction

• Encapsulation

• Inheritance

• Polymorphism

Data Formats, SQLite & Problem Solving

• SQLite CRUD

• JSON, XML, PDF and CSV

• HackerRank practice

• CodeChef practice

Mathematical Computing with NumPy

• Scalars, vectors and matrices

• Central tendency

• IQR

• Variance and standard deviation

• Covariance and correlation

• Skewness and kurtosis

• Central Limit Theorem (CLT)

• Hypothesis tests

• Confidence intervals

• Probability distributions

• Bayes’ theorem

• Marginal, joint and conditional probabilities

Hands-on Milestone Project

SQL & Statistical Analysis Case Study

• Build SQLite CRUD workflows.

• Perform statistical hypothesis tests on Kaggle datasets.

• Interpret statistical test results.

MODULE 2: Data Analytics, Visualization & BI

Learning Objective: Develop practical skills to collect, clean, transform, analyze and visualize data using Python, Pandas, Matplotlib, Seaborn, MySQL and Power BI to generate meaningful business insights.

Data Analysis with Pandas

• Series and DataFrames

• Data loading and inspection

• Data selection, indexing and iteration

• Data cleaning and transformation

• Handling missing values and duplicates

• Sorting and filtering

• Applying functions

• GroupBy and aggregation

• Pivot tables and binning

• Joining and merging datasets

• Data generation and practical exercises

Exploratory Data Analysis (EDA)

• Understanding dataset structure and data quality

• Univariate analysis

• Bivariate analysis

• Descriptive statistics

• Distribution analysis

• Identifying patterns, relationships and anomalies

• Practical EDA using real-world datasets

Data Visualization with Python

• Matplotlib fundamentals

• Seaborn statistical visualization

• Bar charts and line charts

• Histograms and pie charts

• Scatter plots and box plots

• Heatmaps

• Pair plots and distribution plots

• Titles, labels, legends and plot formatting

• Choosing suitable visualizations for business questions

Business Intelligence with Power BI

• Power BI Desktop fundamentals

• Connecting to data sources and loading data

• Power Query for data cleaning and transformation

• Data modelling and relationships

• Creating reports and dashboards

• Interactive visualizations

• Selecting KPIs

• Presenting business insights

MySQL & Database Integration with Python

• Relational database fundamentals

• Creating databases and tables

• Data types

• Primary keys and foreign keys

• SQL CRUD operations

• SELECT, WHERE and ORDER BY

• Data filtering

• Aggregate functions and GROUP BY

• Joins and subqueries

• Practical SQL queries

• Connecting Python to MySQL

• Executing SQL queries from Python

• Integrating MySQL with FastAPI for basic data access

Integrated Data Analytics & Business Reporting

• End-to-end data analysis workflow

• Data preparation using Pandas and SQL

• Exploratory analysis and visualization

• Converting analytical findings into business insights

• Creating a Power BI dashboard from analysed data

• Writing conclusions and presenting findings

Project: Kaggle EDA & Power BI Business Dashboard

• Clean and transform data using Pandas.

• Perform exploratory data analysis.

• Create visualizations.

• Use MySQL for data analysis.

• Create Power BI KPIs and dashboards.

• Present business insights and recommendations.

MODULE 3: Machine Learning & Model Lifecycle

Data Preprocessing & Feature Engineering

• Simple imputation

• KNN imputation

• Iterative imputation

• Label Encoding and One-Hot Encoding

• Outlier detection using IQR, Z-score and LOF

• Min-Max scaling

• Standard scaling

• Robust scaling

• Box-Cox transformation

• PCA, SVD and LDA

• Data leakage

• Scikit-learn workflows

Supervised Machine Learning

• Linear Regression

• Multiple Linear Regression

• Ridge Regression

• Lasso Regression

• Logistic Regression

• K-Nearest Neighbors (KNN)

• Naive Bayes

• Decision Trees

• Support Vector Machines (SVM)

• Voting

• Bagging

• Random Forest

• Boosting

• Stacking

• XGBoost

• Algorithm selection

Unsupervised Learning & Specialized Algorithms

• K-Means clustering

• Hierarchical clustering

• DBSCAN

• Distance and similarity measures

• Apriori

• FP-Growth

• Recommendation systems

• Anomaly detection

• Imbalanced data

• Clustering and pattern discovery

Model Evaluation & Validation

• MAE, MSE, RMSE and R²

• Accuracy

• Confusion Matrix

• Precision, Recall and F1-score

• ROC/AUC

• K-Fold cross-validation

• Stratified K-Fold cross-validation

• Bias-variance

• Overfitting and underfitting

• Learning curves

• Validation curves

• Calibration and threshold tuning

Pipelines & Hyperparameter Optimization

• Machine learning pipelines

• Grid Search

• Random Search

• Bayesian optimization

• Scikit-learn Pipelines

• ColumnTransformer

Model Interpretability & Selection

• Feature importance

• SHAP

• LIME

• Model comparison

• Reproducible experiments

• Systematic model selection

Time Series & Forecasting

• Time series fundamentals

• Temporal data preparation

• ARIMA

• Time-aware train/test split

• Time series evaluation

• Complete pipelines

• Baseline versus optimized models

• Documenting performance and limitations

• Deployment preparation

Project: End-to-End Predictive Machine Learning

• Customer churn prediction

• Employee attrition prediction

• Insurance cost prediction

• Disease classification

Project Deliverables

• Data cleaning and preprocessing

• Baseline model comparison

• Model evaluation and cross-validation

• Hyperparameter tuning

• Model interpretation using SHAP

• Building a machine learning pipeline

• Documenting results, limitations and recommendations

MODULE 4: Deep Learning, Computer Vision & NLP

Deep Learning Fundamentals

• Neurons and perceptrons

• Multi-Layer Perceptron (MLP)

• Forward propagation and backpropagation

• Sigmoid, tanh, ReLU and Softmax

• Loss functions and optimizers

• Gradient descent

• Epochs and batches

• Learning rate

• Overfitting

• Regularization and dropout

Deep Learning Frameworks & Architectures

• TensorFlow/Keras

• PyTorch

• Model training and evaluation

• Saving and loading models

• CNN

• RNN

• LSTM

• GRU

• Transfer learning

• Data augmentation

• Introduction to Hugging Face Datasets and Trainer

Computer Vision with OpenCV

• Image representation using arrays and channels

• Reading and displaying images

• Image resizing and transformations

• Colour spaces

• Thresholding and filtering

• Edge and contour detection

• Feature detection and matching

• Face detection

• Image preprocessing

• Practical exercises

Natural Language Processing (NLP)

• Text preprocessing and normalization

• Tokenization and sentence segmentation

• Stemming and lemmatization

• POS tagging

• Text similarity

• Bag of Words

• TF-IDF

• Text classification

• Sentiment analysis

• Topic modelling

• Word2Vec

• GloVe

• NLTK

Transformers & Attention

• Limitations of sequence models

• Attention and self-attention

• QKV

• Positional encoding and embeddings

• Encoder and decoder

• Multi-head attention

• Feed-forward networks

• Pretrained models

• Hugging Face Transformers

• BERT fundamentals

• GPT fundamentals

Transfer Learning & Fine-Tuning

• Pretrained model selection

• Text classification

• Image classification

• Fine-tuning

• Train, validation and test datasets

• Model evaluation

• Model comparison

• Integrating vision and NLP

Project: Computer Vision & NLP Intelligence System

• Image classification or face/object detection

• Image preprocessing and transfer learning

• Text sentiment analysis or topic classification

• Model comparison using appropriate evaluation metrics

• Demonstration through a notebook or application interface

• Documentation of workflow, results, limitations and improvements

MODULE 5: Generative AI, Fine-Tuning, RAG & AI Applications

Learning Objective: Enable students to build, evaluate and improve Generative AI applications using LLMs, prompt and context engineering, embeddings, retrieval systems, fine-tuning techniques and agentic frameworks.

Generative AI & Large Language Models (LLMs)

• Generative AI fundamentals

• LLM architecture overview

• LLM capabilities and limitations

• Hallucinations and responsible use

• Temperature, top-p and top-k

• Gemini, Grok, Ollama, Llama, Mistral and Qwen

• Running local LLMs using Ollama

• Text, image and video multimodal AI fundamentals

• Model selection based on task, cost, privacy and performance

Prompt Engineering & Structured Outputs

• Zero-shot and few-shot prompting

• Instruction-based and role-based prompting

• Chain-of-thought concepts and task decomposition

• Prompt templates and reusable prompts

• Constraints, examples and output formatting

• Structured outputs using JSON and Pydantic

• Prompt testing, refinement and evaluation

• Practical prompting exercises using real-world tasks

Context Engineering & LLM Application Design

• Context engineering concepts and importance

• Tokens, context windows and context limits

• Context construction and information selection

• Conversation history and memory management

• Context filtering, compression and optimization

• System instructions, user context and retrieved information

• Managing long conversations and relevant context

• Context-aware question answering

• Context quality, security and reliability

LLM Fine-Tuning & Parameter-Efficient Adaptation

• Pretraining, instruction tuning and fine-tuning concepts

• Transfer learning for language models

• Full fine-tuning versus Parameter-Efficient Fine-Tuning (PEFT)

• LoRA (Low-Rank Adaptation)

• QLoRA (Quantized LoRA)

• Dataset preparation and formatting for fine-tuning

• Training configuration and evaluation

• Selecting between prompting, RAG and fine-tuning

Embeddings, Vector Databases & RAG Engineering

• Embeddings and semantic representation

• Cosine similarity

• Euclidean (L2) distance

• Dot product

• Document loading, chunking and metadata

• Chroma, FAISS and Qdrant

• Retrieval-Augmented Generation (RAG) architecture

• Similarity search and top-k retrieval

• Metadata filtering

• Hybrid search using BM25 and vector search

• Context-Augmented Generation (CAG) concepts

• Reranking using bi-encoders and cross-encoders

• Retrieval quality, relevance and grounding

• RAG evaluation using faithfulness, answer relevance, retrieval quality and RAGAS

LLM Application Development, AI & Orchestration

• LangChain fundamentals and reusable components

• ChatPromptTemplates, models, output parsers and retrievers

• Building end-to-end RAG applications

• LangGraph fundamentals and workflow orchestration

• Tool calling and external tool integration

• AI application workflows and task execution

• Conversation memory and persistent storage

• Integrating RAG, structured outputs and tools

• Evaluating application quality, reliability and failure cases

• Building context-aware AI assistants for business use cases

Project: Enterprise Multi-Document RAG & Agentic AI Assistant

• Build a business-oriented AI assistant that answers questions from multiple documents.

• Retrieve relevant evidence.

• Maintain appropriate conversation context.

• Use tools when required.

Project Deliverables

• Load and process multiple documents using document loaders and chunking.

• Generate embeddings and store them in FAISS or another supported vector database.

• Implement semantic retrieval and, where appropriate, hybrid search and reranking.

• Generate context-grounded answers using an LLM, preferably with a local Ollama option for classroom practice.

• Add conversation history, relevant context and persistent storage.

• Integrate a LangChain or LangGraph workflow with at least one useful tool.

• Evaluate retrieval and answer quality, including faithfulness and relevance.

• Document limitations, failure cases and improvements.

MODULE 6: Forward Deployed AI Engineering (FDE) & MLOps

Learning Objective: Enable students to translate real-world business requirements into working AI applications, develop APIs and user interfaces, integrate ML/GenAI systems with databases and external services, and deploy, monitor and maintain applications using modern software engineering and MLOps practices.

Introduction to Forward Deployed AI Engineering

• FDE vs AI Engineer vs ML Engineer

• Business problem identification

• Requirement gathering

• Solution architecture

AI Application Development & Integration

• Python application structure and modular development

• HTML and CSS fundamentals for basic interfaces

• Building applications using Flask

• Rapid AI interfaces using Streamlit and Gradio

• FastAPI fundamentals and REST API development

• Request and response validation using Pydantic

• Asynchronous API concepts

• Integrating SQL databases and external APIs

• Web data extraction using Requests and BeautifulSoup

• Error handling, logging and application configuration

• API testing using Postman

Production-Oriented GenAI & ML Integration

• Integrating trained ML models into Python applications

• Serving model predictions through APIs

• Integrating LLMs through hosted APIs or local Ollama

• Connecting RAG pipelines to business applications

• Integrating embeddings, vector databases and retrievers

• Tool calling and AI workflow integration

• Connecting applications with MySQL and other data sources

• Structured outputs and validation

• Managing timeouts, failures and model errors

• Testing end-to-end AI application workflows

Version Control, Containers & Cloud Deployment

• Git fundamentals and GitHub repositories

• Branching, commits and collaboration workflows

• Dependency management and virtual environments

• Environment variables and secrets management

• Docker images, containers and Dockerfiles

• Containerizing FastAPI and AI applications

• Cloud deployment fundamentals

• Deploying applications on Azure or AWS

• Configuration management across environments

• Deployment testing and troubleshooting

MLOps, CI/CD, Evaluation & Production Reliability

• MLOps concepts and the ML lifecycle

• Experiment tracking using MLflow

• Model registration and versioning

• Data and model artifact management

• Automated testing for applications and APIs

• CI/CD fundamentals using GitHub Actions

• AI application evaluation and regression testing

• Monitoring latency, errors and resource usage

• Security, authentication, authorization and access control

• Debugging and performance optimization

• Model and application update workflows

Project: Production-Ready GenAI/ML Business Application

Design and implement an AI application that solves a specific business problem and demonstrate the complete path from requirements to deployment.

Suggested Project Options

• Enterprise document assistant using RAG and an LLM

• Customer support assistant with database integration and tool calling

• ML prediction service exposed through FastAPI

• Business analytics assistant using SQL and natural-language queries

Required Deliverables

• Business problem statement and solution architecture

• Working AI application with a usable interface or API

• Database or external service integration, where applicable

• GitHub repository with setup instructions and documentation

• Docker-based packaging and deployment

• Testing, evaluation, error handling and security considerations

• MLflow for ML or suitable GenAI evaluation records

• Final demonstration of business value, limitations and future improvements