Debasish Dutta

Developed a Cross Sell Propensity Model using Python (PyCaret) and SQL programming to predict customers most likely to take up a Cash Instalment Plan against their existing credit card. The model, hosted in GCP, produced 5.8X uplift in annualized profit after reaching out to top 30% of model recommended customers.

Developed a Demand Forecasting Pilot Framework for 200 top selling SKUs in German market of a world-renowned cereal brand. These LSTM forecasting models are built using SQL and Python programming (Keras) helped the client in reducing the back ordering by 30% and the delivery time by 4 days. Developed a Customer Segmentation Model (K-Means Clustering) and a Cross-Sell Propensity Model (Random Forest) using Python and SQL programming for the top most consumer electronics conglomerate from South Korea. These models helped the client in increasing the average order value in Life Style Buyer segment by 17% and online and offline campaign effectiveness by 35% and 15% respectively. Developed a Customer Migration Predictive Model (XGBoost) using PySpark and Hive programming for the largest telecom operator from Canada. This classification model accurately predicted the customer movement from legacy to IPTV plan thus helped the client in reducing the network congestion by 25% at CMTS level.

Developed a Warranty Claim Classification Engine (NLP, Naïve Bayes Classifier) using Python programming for one of the largest aircraft engine manufacturers from North America. This Model helped the client in reducing claim adjudication time by 20 days and FTE count by 7 claim assessors. Developed Condition Based Asset Maintenance Models (Logistic Regression) using Python and SQL programming for one the largest oil and gas company from Australia. These models accurately predicted the state change within the asset that can leads to equipment breakdown resulting the decrease in unscheduled asset maintenance cost by 35%. Improved Hot Blast Temperature by 140C using Unsupervised Clustering Technique (K-Means) in Python Programming for the largest steel manufacturer in India. Operating regimes for most crucial process parameters suggested by the analytics model yielded in to estimated ~23 crore INR savings in coke consumption.

Developed an End-To-End Claim Volume Forecasting Framework (Time Series Classification Engine & Automated Forecasting Machine) in R Programming for the 220 claim processing units. This framework helped the business in improving claim adjudication time by 23% and cross training activities by 55%. Developed a Complaint Tracking Module (SMOTE, Random Forest Model) in MS-SQL and R Programming which resulted in to 24% reduction in complaints for Medicare Advantage insurance plans within the span of three months. Subsequently within a year of time, STAR ratings of these insurance plans improved by 0.3 on the scale of 5 due to lower complaints and appeal and higher customer satisfaction.

Developed multiple Time Series Forecasting Models for accurate capacity planning of offshore as well as onshore Hedge Fund Claim Processing units. Acted as Primary Point of Contact for multiple CQI (Continuous Quality Improvement) and KPI (Key Performance Indicator) dashboards.