Microsoft Accelerator Program in Data Science and Artificial Intelligence

This week provides a comprehensive introduction to the key terminology in the fields of Data Science and AI, and helps participants identify real-world business applications of data science and AI across various industries.

Topics Covered:

• The role of data in the life of the general population 
• Paradigms in data science: Inferential and computational 
• Key Terminology (AI, DL, ML, GenAI, LLM) 
• Business Applications of Data Science and AI

Sample Case Study: Airline Customer Review Sentiment Analysis Analyze the reviews in tweets shared by customers of an airline regarding their experience with the services provided, and build a predictive model that can parse the customer reviews and predict the sentiment of the customer

This week provides a comprehensive introduction to the core Python programming constructs such as variables, data structures, conditional statements, loops, and functions. Participants will also learn how to load, manipulate, transform, and save datasets using Numpy and Pandas to support basic data analysis workflows.

Topics Covered:

• Google Colab 
• Variables, Datatypes, and Data Structures 
• Conditional and Looping Statements 
• Functions 
• NumPy arrays and function 
• Pandas Series and DataFrames (Creating, Accessing, Modifying, and Combining) 
• Pandas Functions Saving and loading datasets using Pandas

This week provides a comprehensive coverage of univariate, bivariate, and multivariate analysis using appropriate visualization techniques to uncover patterns and relationships. Participants will also explore ways to identify and handle missing values and outliers using standard data preprocessing techniques to ensure data quality.

Topics Covered:

• Data overview 
• Univariate analysis (Histogram, Boxplots, and Bar graphs) 
• Bivariate/Multivariate analysis (Line Plot, Scatterplot, Heatmaps, and Pairplots) 
• Missing value treatment Outlier detection and treatment

This week will cover the fundamental database concepts and the role of SQL in querying structured data. Participants will learn how to write SQL queries to fetch, filter, and aggregate data from relational databases, and use SQL in-built functions to transform and analyze data effectively.

Topics Covered:

• Introduction to Databases and SQL 
• Fetching data 
• Filtering data 
• Aggregating data 
• In-built functions (Numeric, Datetime, Strings)

This week focuses on constructing SQL queries using various types of joins to combine data across multiple tables, implementing window functions to perform advanced analytics like ranking, and writing subqueries to retrieve and filter data in nested contexts for complex query logic.

Topics Covered: 

• Joins 
• Window functions 
• Subqueries
• Order of Execution

In this week, participants will analyze relationships between variables using correlation and assess their suitability for regression modeling, build and interpret simple and multiple linear regression models, including handling categorical predictors, and evaluate regression models using appropriate performance metrics to assess prediction accuracy.

Topics Covered: 

• Business Problem and Solution Space - Regression 
• Correlation and Linear Relationships 
• Simple and Multiple Linear Regression 
• Categorical Variables in Linear Regression 
• Regression Metrics

This week will cover the structure and working of decision trees for classification and regression tasks. Participants will understand how to apply impurity measures and splitting criteria to grow effective decision trees, implement pruning techniques to control overfitting, and evaluate decision tree models using appropriate performance metrics in business scenarios.

Topics Covered: 

• Business Problem and Solution Space - Classification 
• Introduction to Decision Trees 
• Impurity Measures and Splitting Criteria 
• Pruning 
• Decision Trees for Regression

This week will cover the business scenarios suitable for clustering and formulate them as unsupervised learning problems. Participants will compare different distance metrics and assess their impact on clustering outcomes, differentiate between various clustering techniques, and implement K-means clustering to segment data and interpret results to generate business insights

Topics Covered:

• Business Problem and Solution Space - Clustering 
• Distance Metrics 
• Introduction to Clustering 
• Types of Clustering 
• K-means Clustering

In this week, participants will learn how to apply K-fold cross-validation to assess model robustness and reduce overfitting, implement resampling techniques like oversampling and undersampling to handle class imbalance in datasets, and tune model hyperparameters using GridSearchCV and evaluate performance improvements. Participants will also learn how to deploy machine learning models using Streamlit to create user-friendly applications that support business decision-making.

Topics Covered: 

• K-fold cross-validation 
• Oversampling and Undersampling 
• Hyperparameter Tuning with GridSearchCV 
• Introduction to Model 
• Deployment Deployment using Streamlit

Sample Case Study: Machine Predictive Maintenance Analyze the data of an auto component manufacturing company and develop a predictive model to detect potential machine failures, determine the most influential factors on machine health, and provide recommendations for cost optimization to the management.

This week will cover the role of neural networks in solving complex problems, differentiate between ANN and CNN architectures, identify suitable use cases for each in business contexts, and build and train basic MLP and CNN models using appropriate architectures and optimization strategies. Participants will also explore the concept of transfer learning and demonstrate how pre-trained models can be adapted for new tasks.


Topics Covered:

• Deep Learning and History 
• Multi-layer Perceptron 
• Types of Activation Functions 
• Training a Neural Network 
•  Backpropagation 
• Types of Optimizers 
• ANN Vs CNN 
• CNN Architecture 
• Introduction to Transfer Learning
  
  

Sample Case Study: COVID Detection Build an image classifier using pre-trained CNN architectures for COVID detection, enabling efficient and scalable medical diagnostics from imaging or clinical data.

This week will cover the core concepts of Generative AI and Large Language Models (LLMs), analyze the architectural components that power ChatGPT and their role in generating human-like responses, and define the concept of prompt engineering. Participants will also apply different prompting techniques to optimize LLM responses for specific business use cases.

Topics Covered:

• Introduction to Generative AI 
• History of Generative AI 
• Overview of Large Language Models (LLMs) 
• ChatGPT - Overview and Business Applications 
• Breaking Down ChatGPT 
• Introduction to Prompt Engineering 
• Strategies for Devising Effective Prompts

Sample Case Study: Restaurant Review Analysis Analyze the provided sample of customer reviews for different restaurants for a leading global food aggregator and build a predictive model to analyze the reviews and tag them, thereby enhancing the company's ability to understand customer sentiments at scale, enabling data-driven decision-making, and improving overall customer satisfaction.

This week will cover the core principles of Retrieval-Augmented Generation (RAG) and its significance in enhancing LLM performance for domain-specific tasks. Participants will gain an understanding of the key components of a RAG pipeline, including retrievers, vector databases, and generators, demonstrate the use of semantic search and vector databases to retrieve relevant context for generative tasks, and design and evaluate prompts tailored for RAG workflows to improve output quality and business task alignment

Topics Covered: 

• Introduction to Retrieval Augmented Generation (RAG) 
• Building Blocks of RAG 
• Vector Database and Semantic Search 
• Devising and Evaluating Prompts for RAG

Sample Case Study: Real-Time Inventory Tracking & Logistics Optimization Integrate external knowledge sources to enhance data retrieval accuracy and implement Retrieval-Augmented Generation (RAG) to deliver real-time inventory and logistics updates, ensuring seamless supply chain management and proactive decision-making