Nitin Kshatriya

Developed and implemented pricing solution in production with end-to-end pipelines to simulate the impact of price change on forecasted SKU sales for Philip Morris International's tobacco portfolio for over 15 markets with 95% model accuracy. Modelled the impact of price changes on volume and used the log-loss model to evaluate the impact of price on volume, which allowed us to calculate the elasticity of each SKU. Utilized a blend of Moving Average and ARIMA models to generate base forecasts for each SKU, enabling simulations of price change effects on volume while considering SKU elasticity. Project Impact: Resulted in optimized pricing strategies and sales forecasts, contributing to data-backed decision-making and market competitiveness across diverse regions.

Leveraging OCR, Object Detection, and Named Entity Recognition technologies, streamlined the KYC process, eliminating manual data entry, and significantly improving user experience. Managed end-to-end implementation from requirement gathering, data collection, annotation, cleaning, training, and deployment of Object Detection and Named Entity Recognition models. Developed and trained custom models using TensorFlow and Spacy. Deployed using Docker and TensorFlow Serving. Project Impact: Enhanced user experience by eliminating manual data entry which improved data accuracy and processing time.

Led end-to-end implementation of accurate garbage categorization model using TensorFlow and Docker for real-time deployment, supporting Intel's SwachhMap waste management initiative. Played a key role in processing data from water level sensors using pandas, enabling real-time water usage prediction, and actively contributed to decision-making on data analysis for users' insights and experience enhancement for Waltr.in. Successfully extracted data from Blood Pressure Machine screens using CV2 and classification models, enabling seamless integration of data for effective health monitoring.