Bhupesh Hada

Graduate Student Assistant for the course EAS509: Statistical Learning and Data Mining 2 offered by the SEAS department under Professor Nikolay Simakov

Graduate Student Assistant for the course CDA541: Statistical Learning 1 offered in the Master of Professional Studies in Data Science and Applications program under Prof. Rachael Hageman Blair and Prof. Han Yu

● Developed an AI-powered eCommerce Advertising solution that improved the Amazon ad performance and led to 60% increment in the purchase rate. ● Created XGBoost model to identify best-performing line items for a specific campaign. Further, launched a campaign monitoring process to keep track of campaign performance. ● The solution proved vital in cost optimization and resulted in a 10x improvement in ROAS.

EU Price Optimization | Azure Data Lake, Databricks, Pyspark, Python, Bayesian Regression, PyGAD ● Built an ensemble of Genetic Algorithm, Particle Swarm Optimization and Differential Evolution optimization models against the objective function dealing with KPIs like Turnover, Risk, Gross Profit, Gross Margin to get the best product prices. ● Led a team of 4 in implementing Bayesian Regression models for product - level price elasticities and automated the process flow through data pipelines on Microsoft Azure. ● Recommended product prices that successfully reduced client’s revenue loss due to arbitrage. Edge Analytics | Pyspark, Azure Blobs, Azure Data Lake, Azure Data Factory, Azure SQL Database, Azure Analysis Services, Power BI ● Designed and deployed a product on Azure cloud to track and analyze the performance of the promotions offered to the retailers with the sell-in. ● Developed ETL workflows using Azure cloud components and configured SQL data warehouse for optimal storage which was subsequently used to load data in Power BI reports. ● This solution cut down the time taken to run the process by almost 50% with zero manual intervention. Salesman Incentive Optimization | R, gtools, ggplot, dplyr ● Designed an automated tool that orchestrated data preprocessing, collating a wide range of salesman performance data, and subsequently running more than 20,000 regression models against the business-defined targets. Used stepwise regression to identify the top contributing features/KPIs. ● Recommended primary and secondary KPIs with the graphical visualizations and the corresponding optimal values based on the inflection point in each of the KPI plot.

● Built a LightGBM model to identify the cars that can be converted (sold to the local dealers), for a B2C firm that deals in secondhand cars. ● Leveraged the model's conversion prediction probabilities to categorize cars into 3 distinct groups based on the conversion percentages. This classification proved pivotal in prioritizing and crafting a targeted marketing strategy aimed at maximizing conversions in the high-percentage category.

Organize sports events for the Physics Departement of IIT Roorkee

The aim of this research work is to study the various factors which affect the tourist inflow in various states of India.This study involves collecting and processing the views from the blogs of the people(Sentiment Analysis) who have been to various states of the country and have mentioned their experiences in their blogs.

A model has been trained on a number of computer generated and handwritten invoice pictures in order to detect the filled values and the unfilled values in an invoice. values could be any name(alphabatic) or a number in an invoice https://imgur.com/a/cX1Rh (what we actually want)

I along with a couple other students have worked in the direction of exploiting the spectrum of inorganic scintillators (LaBr3, NaI) in order to calculate the W-value of scintillator that will help in understanding resolution of scintillators.