Assortment Pack Optimization

Paper submitted to Annals of Operations Research.

Presented this work at INFORMS 2023.

Developed an application to optimize the design and allocation of assortment packs for apparel, ensuring store-level demand is met efficiently. Built a hybrid solver combining k-means clustering, simulated annealing, linear programming, and greedy heuristics. The method, framed as a Bi-Linear Integer Program, improved in-season revenue and reduced computation time.

Keywords: Ensemble Optimization | K-means | Simulated Annealing | Linear Optimization | Greedy Heuristics | Property Testing | Stakeholder Engagement

Optimizing 3PL-to-Target Transfers

Developed a Streamlit app and enhanced the allocation algorithm to improve transfer efficiency. Led two data scientists and collaborated with cross-functional teams, influencing strategic decisions. This project reduced manual workload, increased trailer fill rate, and cut out-of-stocks.

Keywords: Inventory Management | Streamlit Application | HDFS & Spark | Cross-functional Partnership | Leadership

Predicting Item Resale Values for Reverse Logistics

Developed predictive models to estimate resale values of items sold through a local D2C channel, leveraging EDA and feature engineering. Built hierarchical averaging models and applied regression techniques, including LASSO, Random Forest, and XGBoost. Achieved reduction in Mean-Squared Error using Random Forest compared to the baseline model, enhancing pricing accuracy.

Keywords: Exploratory Data Analysis | Machine Learning | Random Forest | XGBoost

Personalizing Outfitting Recommendations

Developed a system to personalize recommendations for complementary products based on user searches. Implemented multi-armed bandit algorithms to optimize recommendations. The Lin-UCB bandit algorithm achieved the highest improvement in CTR over a non-personalized random policy, enhancing user engagement.

Keywords: Personalization | Recommendations | Multi-armed Bandits | Contextual Bandits | Lin-UCB | Reinforcement Learning

Persuading Customers to Buy Early: The Value of Personalized Information Provisioning

K. Drakopoulos, S. Jain, R. Randhawa

Published in Management Science | Paper

We study the pricing & information provisioning game of a seller who is (ex-post) better informed about product availability. Using a Bayesian persuasion framework, we find that public information provisioning has limited value while personalized information provisioning is profitable, having attributes like personalized pricing.

Effects of User Voting and Reputation on Fact-Checking

K. Drakopoulos, S. Jain, R. Randhawa

We consider a platform with two users who can fact-check a common article to reduce their misinformation. We examine the effects of introducing a voting mechanism in which each user can upvote or downvote the article to improve its public reputation.

Searching for an Infection in a Network

K. Drakopoulos, S. Jain, R. Randhawa

We analyze a graph in which an infection spreads deterministically from an unknown infected node to its neighbors at each time step. Our goal is to design a search policy that finds an infected node in minimal time. We formulate the problem as an equivalent graph covering problem and develop an integer linear program for the same. We show that the problem is NP-Hard for an arbitrary graph.

Operations Management (BUAD311), USC Marshall

Role: Instructor

Received Marshall PhD Teaching Award 2020

Course | Syllabus | Student Evaluation

Taught a class of 57 undergraduates.

Topics Covered: Business Process Management | Capacity Management | Queueing Theory | Optimization | Revenue Management | Inventory and Supply Chain Management | Forecasting

Operations Management (BUAD311), USC Marshall

Role: Teaching Assistant

Course

Served as the only Teaching Assistant for 600+ undergraduates.

Topics Covered: Business Process Management | Capacity Management | Queueing Theory | Optimization | Revenue Management | Inventory and Supply Chain Management | Forecasting

Introduction to Operations Research (MCL261), IIT Delhi

Role: Teaching Assistant

Served as the only Teaching Assistant for 150+ undergraduates.

Topics Covered: Linear Programming | Simplex Method | Revised Simplex | Duality | Sensitivity Analysis | Transportation Problems | Probability Distributions | Queueing Theory | Markov Chains