During my Summer Intern role at IDEAS-TIH (ISI Kolkata), I compared six deep learning model variants in Python and Keras, balancing accuracy against computational cost. I streamlined a model’s design to cut its computational footprint by 81%, then redesigned it to bring overall resource use down 72% with minimal quality loss.
My economics projects have applied data analysis to credit risk, public health schemes, and wealth inequality. For my CreditCardDefaultRisk Modeling project, I tested logistic regression models and recommended the simpler model for production scoring when added interaction terms improved prediction accuracy by only 0.21 percentage points.

