Data Scientist
CurrentResponsible for analyzing large sets of data to draw conclusions and make recommendations to the business using statistical models. Involved in discussions with business teams to understand the business process mapping and came up with business use cases to improve sales, predictive product market penetration using ML models. Worked closely with project managers to develop and expand business use cases ideas for implementations. Used XGBoost model to identify the… Show more Responsible for analyzing large sets of data to draw conclusions and make recommendations to the business using statistical models. Involved in discussions with business teams to understand the business process mapping and came up with business use cases to improve sales, predictive product market penetration using ML models. Worked closely with project managers to develop and expand business use cases ideas for implementations. Used XGBoost model to identify the features that predict the sales metrics and KNN algorithm to generate a comparable set of accounts that behave similarly. Used AWS S3 bucket to load the data from SQL database. Worked in development of product penetration model enhancements that is deployable and schedulable in AWS sage maker and Amazon Code Pipeline. Created End to End ML models based on State rules to handle unique rules around account purchase patterns and regulations. Worked in generalizing model to streamline penetration evaluation for a user driven product. (Brands, Sku groups, Product types etc.) Lead the evaluation and inclusion of appropriate affinity metrics in penetration based on region and account demographics. Deployed 10+ models of product penetration models in AWS cloud platform by implementing diverse Amazon pipelines catering to multiple regions. Leveraged ChatGPT API to process audio, video, text files and generated themes, summary, sentiment score and weekly customer sentiment reports for customer service center directors. Involved in fast paced & complex demand sensing program to acquire data from 20+ external sources and 25+ internal sources and then cleansed data, scaled features, reduced dimensions then ingesting data sets in into data bricks environment. Crafted models to accurately predict (less 7% variance) the demand for upcoming 12 months including safety stock and dead stocks. Derived specific features based on feature importance which contributes to dead stocks and eventually reduction in it. Show less