Prabal Das

I have completed my PhD in hydro climatology from IIT (ISM) Dhanbad in Jan 2024. I am currently engaged as a post doc researcher at the University of Texas Arlington, USA.

I am currently engaged in a project sponsored by the National Oceanic and Atmospheric Administration (NOAA). My work involves developing a machine learning-based model specifically designed for hourly forecasting of precipitation classes across the Continental United States (CONUS) region.

My expertise includes

• Proficient in machine learning with a strong academic background, including publications related to ML during PhD. 
• Experienced in using Python and R scripting for developing ML models using scikitlearn, caret, pycaret etc. 
• Developed Deep Learning (LSTM/CNN) models for various applications using Tensor Flow both for regression and classification problem. 
• Well-versed in a variety of other ML algorithms using scikit-learn and caret
• Experienced in data collection, manipulation and analysis from various sources (cloud based or AWS (like GEFS, Reanalysis etc.) also from APIs like mPING using numpy, pickle, rds etc. (NASA’s app for reporting precipitation type) 
• Data visualization using matplotlib, seaborn, ggplot2 etc. 
• Experienced in LINUX or UNIX based bash or shell scripting

I am a PhD graduate from Indian Institute of Technology (Indian School of Mines), Dhanbad in Hydro climatology from the department of Civil Engineering. My thesis was titled “Unveiling Key Drivers in Hydrometeorological Processes: Feature Selection Using Bayesian Networks for Improved Understanding and Prediction”. During my PhD, I focused on:

• Focused on various feature selection algorithms like Bayesian Networks (BN), Recursive Feature Elimination etc., for identifying the important precursors responsible for modeling various primary and secondary hydrologic variables like rainfall and streamflow and for characterization of tertiary variables like drought. 
• Influence of large-scale climate modes on basin scale streamflow and rainfall is also explored. 
• Selection of suitable GCMs for improving the accuracy of GCM outputs