Shashank Vats

PriceSmart

• Formulated model in order to solve for multi-hierarchical targets based on iterative approach and handle more customizable user set constraints.
• Formulated rules to achieve metric targets - margin, sell-through, and revenue based on their order of priority. These targets have to be met for both in-season and clearance periods.
• Formulated rules engine for client-based needs like keeping the ECOM price +-10% that of B&M, and selecting the discounts based on average weeks for item groups.
• Formulated new approach to solve the problem using slack variables in order to reduce the number of optimization process which led to reduction in time by ~40%.
• Modified the model to build multiple versions based on the requirements of various clients.

MarkSmart

• Worked on building model to obtain optimal cadence for each product group/bin for a planning horizon as a Mixed Integer Linear Programming model using Gurobi .
• The model has to achieve high margin while considering sell-through requirements and margin percentages.
• Cadence for each bin has to respect a few constraints such as number of distinct discounts selected for a bin, min/max frequency of selected discount bin, step-size in bins based on previous week's discount, etc.
• Improved solve time of model by ~30% using various techniques like parameter tuning, constraint optimization, neighborhood search, etc.
• Modified codebase so as to enable model to run using LocalSolver and compare how it fares with respect to Gurobi.

AllocateSmart

• Worked on building model to minimize over-allocation/under-allocation of packs from DC to store for a given time horizon using Mixed Integer Linear Programming Model using Gurobi.
• Improved model solve time by 2x using various techniques like converting model from one-step to two-step approach, giving warm start to model, parameter tuning, etc.
• Modified codebase so as to enable model to run using LocalSolver and compare how it fares with respect to Gurobi.

ad-hoc Analysis

• Customer Segmentation for optimization of marketing and help formulate segment-specific strategies.
• Propensity Analysis by building a robust model that can give probability of whether a customer will shop or not for the next campaign.
• Market Basket Analysis to find the association between the items and recommend the item-groups having affinity for the purchase in the same transaction.