I participated in the AI and Data Science Program offered by the MIT Institute for Data, Systems, and Society from April to August 2026. The program focused on using responsible AI, data, and statistical methods to address practical business challenges.
Throughout the program, I worked on projects that involved solving real-world business problems using AI and machine learning techniques. These projects strengthened my ability to use data-driven approaches for business decision-making, develop practical recommendations, and translate analytical findings into actionable business solutions.
Link to certficate: https://www.mygreatlearning.com/certificate/LLYZJNHN
Project 1: Amazon Product Recommendation System
The rapid growth of e-commerce has created an overwhelming number of product choices for consumers. Recommender systems help businesses address this challenge by providing personalized product suggestions based on customer preferences and behavior, improving customer engagement and supporting sales.
Objectives: To use labeled data of Amazon reviews of products to build a recommendation system that helps in recommending products to online consumers.
Data: The Amazon dataset contains information on userid, productid, rating
Machine Learning Methods: Rank-based recommendation systems, collaborative filtering, and model-based collaborative filtering (matrix factorization)
Results: The rank-based model offers a simple way to recommend popular products to new users with no purchase history, addressing the cold-start problem. Collaborative filtering personalizes recommendations by drawing on similarities between users and products. The optimized Singular Value Decomposition (SVD) model proved the most practical for deployment. Although its RMSE was slightly higher than the best item-item KNN model, it can predict ratings for user-product pairs not yet observed, enabling scalable, personalized recommendations. Together, these approaches can increase customer engagement, raise conversion rates, and support cross-selling. Future improvements include hybrid models, content-based features, and deep learning-based recommenders to further enhance personalization and accuracy.
Project 2: Customer Personality Segmentation
Objective: To identify distinct customer segments that enable personalized marketing campaigns, improve conversion rates, strengthen retention of high-value customers, and optimize resource allocation across inventory, pricing, and store layout.
Data: Historical customer data from Retail X covering demographics, spending, purchase and campaign interactions, and shopping behavior.
Machine Learning Methods: Exploratory data analysis and K-means clustering
Results: Segmentation revealed distinct customer groups with different spending patterns and engagement behaviors, enabling targeted rather than one-size-fits-all marketing. For price-sensitive, lower-income customers who visit the website frequently, I recommended online promotions and discounts to boost conversion. For high-value customers, I recommended relationship-focused campaigns emphasizing convenience, trust, and exclusivity to strengthen loyalty and retention. These insights help the retailer direct marketing spend, inventory, and pricing toward the segments most likely to generate returns.
Project 3: Lead Conversion Prediction and Customer Profiling for the EdTech company ExtraaLearn
ExtraaLearn is an EdTech company offering programs in emerging technologies to help students and professionals upskill and reskill. It generates leads through digital channels such as social media, website interactions, brochure downloads, and email inquiries, then nurtures them through sales calls and emails to convert them into paying customers. The key business challenge is identifying the leads most likely to convert, so the company can target sales efforts and allocate resources more effectively.
Objective: To develop a machine learning model that predicts lead conversion and identifies the key factors driving it.
Data: Lead-level data on demographics, digital engagement, marketing exposure, and conversion outcomes. Variables include age, occupation, first interaction channel, profile completion, website visits, time on site, page views per visit, last activity, print and digital advertising exposure, educational channels, referrals, and conversion status.
Machine Learning Methods: Decision trees and random forest
Results: The analysis identified website engagement, profile completion, and age as key drivers of conversion. I recommended prioritizing website engagement, encouraging profile completion through targeted incentives, and tailoring marketing campaigns by age group. I also recommended deploying the random forest model to score incoming leads, so the sales team can focus follow-up on high-potential prospects. This improves conversion rates and makes better use of resources.