Amitabh G.

About Amitabh G.

I have more than 7 years of experience as a data scientist. I have built expertise in various areas including regression, time series forecasting, computer vision, natural language processing, causal inference and Generative AI. I have experience in building AI/ML driven products for forecasting critical variables like cash, revenue, churn.I led clinical research projects at Providence India where I have worked on multiple clinical research projects using EHR data to infer causal effects of medications or infections on multiple health outcomes. Built data pipelines for cohort extraction and data cleaning. I designed and developed the statistical analysis pipeline for hypothesis testing, survival analysis while leveraging Python multiprocessing on Databricks. Poster presentation on COVID-19: designed and developed multiple ML systems for multiple prediction use-cases resulting in cost savings. Delivered use-cases on multiple different types of payment models (capitation, bundled payments), e.g, 1) predicting patient churn to reduce the revenue loss from churn in managed care programs in the US, 2) predicting re-admission probability and length of stay for high-risk stroke patients and others. I developed forecasting models in order to help stakeholders plan and optimize hospital resources, e.g, multivariate time series model to predict net cash payments. Created time-based features, Fourier features to capture multiple seasonality, and performed box-cox transform to handle heterokedasticity in data. Achieved the forecast error of less than 2.5 percent MAPE for real time forecasts and eventually led to very efficient resource planning. I have worked on application layer of Generative AI where I have leveraged LLMs for feature extraction to support research outcomes, developed RAG system on research documents to expedite researchers’ workflow.

Experience

1. Lead Data Scientist

   Providence India

   Apr 2023 — Present · Hyderabad, IN

   Leading clinical research initiatives at Providence India. Conducted research and writing papers on the causal effects of COVID-19 on 1.5 million patients for 780 health conditions. Built the statistical analysis pipeline for hypothesis testing, survival analysis while leveraging Python multiprocessing on Databricks. I have also designed and developed multiple ML systems for multiple prediction use-cases resulting in cost savings. Delivered use-cases on multiple different types of payment models (capitation, bundled payments), e.g, 1) predicting patient churn to reduce the revenue loss from churn in managed care programs in the US, 2) predicting re-admission probability and length of stay for high-risk stroke patients and others. I have developed forecasting models in order to help stakeholders plan and optimize hospital resources, e.g, multivariate time series model to predict net cash payments. Created time-based features, Fourier features to capture multiple seasonality, and performed box-cox transform to handle heterokedasticity in data. Achieved the forecast error of less than 2.5 percent MAPE for real time forecasts and eventually led to very efficient resource planning. I have also been at the forefront of AI where I have leveraged LLMs for feature extraction to support research outcomes, developed RAG system on research documents to expedite researchers’ workflow.

Education

• National Power Training Institute (ER), Durgapur

  Bachelor of Technology - BTech, Engineering

  2012 — 2016

• Indian Institute of Science (IISc)

  Master of Management, Statistics, AI/ML and Management fundamentals