Rupesh Prasad

Leads Prediction: Led the development of a classification model that employs XGBoost to predict the likelihood of a close win or loss for open deals. This information is helping to prioritize lead generation efforts and helping sales teams identify areas for improvement in their sales processes.
Raw Material Pricing Prediction: Developed statistical models that assist in predicting the prices of raw materials such as polyethylene, based on several indexes, such as Brent crude oil and naphtha. This model helps marketing teams make informed pricing decisions based on predicted raw material costs, thereby ensuring profitability and competitiveness.

Eficiency analysis for carbonation process: Reduced variance in the individual tower efficiency and moved towards Golden Efficiency by ~70%.
Turbine Trip Analysis: Developed a classifier to identify the factors affecting the wind turbine trip at a particular time zone. Total number of trips has reduced from 6000 per year to ~2500 per year.
AutoEDA: Developed an automated code which takes a tabular data as input and output an excel having various EDA results like Number of outliers, correlation analysis, descriptive statistics for numeric variable, frequency analysis, Plots(Box plot, bar plot, histogram, scatter plot etc.)
AutoML: Developed an automated code which takes data frame, dependent variable name, problem type as input and does all basic preprocessing before modeling (like outlier/missing treatment), recode categorical variable, create polynomial feature, for top variables, split the data and finally fit multiple models and report validation metric based on problem type.

NLP Claims Analytics [US PCG Claims classification]: Developed an NLP integrated Machine Learning (ML) model that helps to proactively identify PCG Claims handling by "Public Adjusters" & Catastrophe (CAT) claims. This model is helping AIG to identify more than 70% of adjusters handled claim correctly.
Loss Cost Model: Developed loss cost models (or pure premium) for travel insurance product for various APAC regions.
Conversion Model: Built a model to identify drivers of travel insurance purchase, thereby improving conversion rates and underwriting profits. Achieved a 20% increase in conversion due to the model.

Next Best Offer: Used Latent Markov model as the underlying statistical tool to make the right offer to the right customer at the right point in time. Achieved a 5%-25% overall increase in sell.

Product Propensity Models: Predicted the likelihood of a customer purchasing a particular product. Helped product managers optimize email send frequency and offering discounts to enhance the chance of product take up.

Customer segmentation :- Conducted customer segmentation to optimize business planning and marketing decisions resulting in overall improvement in business performance.
Attrition analysis : Completed the attrition analysis project, which reduced the attrition rate by 35% and enabled the launch of pro-active customer retention campaigns.
Pure premium modelling : Developed generalized linear models for pure premium modeling of bodily injury coverage, enabling clients to determine the pure premium for all customers using primary variables.