Dec 2021 – Dec 2023 · 2 yrs · Bengaluru, Karnataka, India

• Deployed an enterprise level demand forecast solution with capabilities of transfer learning and feature store in a reliable, scalable and secure way using containerized orchestration on Azure kubernetes for a leading alcoholic beverage company. It is helping the organization in easy prototyping of time series forecasts and streamlining fragmented forecasting solutions in multiple markets.
• Institutionalized a platform agnostic framework for model monitoring in production which reduced the monitoring cost by 30%.
• Engineered a product recommendation system using a hybrid approach of Behaviour Sequence Transform (BST) model and content based recommendation.
• Overhauled outlet segmentation and deployed on Azure Databricks with REST API endpoint. This led to reduction of “cost to serve” by 15% driven through optimization of picking and transportation cost in a 3 months A/B test.

Jan 2021 – May 2021 · 4 mos · Hyderabad, Telangana, India

• Upgraded and automated the Intent Analysis tool leveraging the power of LLMs and Social Media to understand Mood of the Nation. It helped in better targeting to woo the voters in Bengal Election 2021.
• Conducted survey using questions based on personality trait and hypothesized it to calculate expected vote share in the upcoming election 2021.
• Analyzed the past election data to infer relation among caste, religion and vote percentage.

Feb 2019 – Dec 2020 · 1 yr 10 mos · Hyderabad, Telangana, India

• Productionized Anomaly Detection module operated on time series such as S&P Dow Jones Indices and sales data of one of the leading retail stores of the US.
• Designed an Explanation Pipeline using Shapley values for stock market data and generalized it for any time series data to show the extent of features/covariates on target value.
• Strategised Dynamic Pricing using traditional as well as advanced reinforcement learning method, Deep-Q Learning on simulated e-commerce data.
• Pricing strategy like- Price discrimination to extract consumer surplus
• 2nd degree price discrimination like block pricing (Created demand schedule from different prices offered at various times)
• 3rd degree price discrimination: Using price elasticity in different market
• Identification of extent of complementary goods and substitutes cross selling.
• Productionized Anomaly Detection module operated on time series sales data.
• Assembled Data Engineering pipeline using Logstash and streamlined features for downstream consumption in modeling and visualization. Supervised the end to end pipeline through Kibana Dashboard.
• Customer segmentation for brand promotion using clustering approach with attributes like volume of transaction, number of products, category of products, active in last 15 days etc.

Jul 2017 – Feb 2019 · 1 yr 7 mos

• Applied Timeseries forecasting models like exponential smoothing, Holt's Winter, ARIMA, Seasonal ARIMA on NOKIA Dataset in R language to predict sales of several products for next 6 months with 80% accuracy.
• Applied Multipe Linear Regression with regularization on CARGLASS EMEA region Dataset in Python to predict sale of glass for next 3 months with 82% accuracy.
• Applied Logistics Regression and SVM on CARGLASS dataset with metrics like Precision, Recall and AUROC to predict which glass (Rear/ Windshield) a customer is supposed to replace.
• Segmented customers of DELL into four clusters using KMean clustering algorithm in Python with features likes RAM, HD, Screen Size etc.
• Did several POCs for customer like Coca Cola and Fitbit.

Jul 2016 – Jun 2017 · 11 mos

• Worked on Replenishment Engine in JAVA.