Summary
Overview
Work History
Education
Skills
Timeline
Generic

Soham Das

Bangalore

Summary

Senior AI/ML Associate at JP Morgan Chase with expertise in Credit Risk model development and stakeholder collaboration. Proficient in machine learning techniques such as XgBoost, leading to improved collections strategies and optimized account management. Demonstrated ability to deliver impactful solutions that enhance business outcomes while maintaining model reliability and compliance.

Overview

10
10
years of professional experience

Work History

Senior AI/ML Associate

JP Morgan Chase India Pvt Ltd
Bengaluru
12.2019 - Current
  • Responsible for end-to-end development, maintenance of Credit Risk models for Chase NA credit card portfolio, used for optimization of Collections and Account Management Strategies.
  • Worked closely with Card Strategy, Model Governance and Tech teams to ensure timely delivery of models serving business purpose while adhering to model risk and controls.
  • Developed a regression model using XgBoost technique to predict customers reported Income.
  • Business uses Debt to Income (DTI) ratio in different account management strategies as a proxy of customers' ability to pay.
  • The predicted income gets used as a part of available sources of income to calculate DTI.
  • Created a framework to establish reliability of reported income which was used as target variable.
  • Development included variable selection, hyper parameter tuning and using customized asymmetric loss function to penalize the over prediction of income more than under prediction.
  • Developed a classification model using XgBoost technique to identify among the delinquent Chase cards customer base, which customers though have ability to pay, chooses to pay competitors.
  • Identifying those customers will help in offering the right customers the right payment programs and increase collections.
  • Development included target variable creation by roll rate analysis, variable selection and hyper parameter tuning.
  • Validation was done in both in time and out of time samples.
  • Business wanted to redevelop one of the collection models, which used to predict the likelihood of a delinquent Chase credit cards going into charge off.
  • Conducted EDA to establish the fact the bucket 1 and 2 accounts show different payment patterns than the higher buckets and suggested two different models: one for bucket 1 and 2, and one for higher buckets.
  • Developed a classification XgBoost model to identify that among the bucket 1 and bucket 2 accounts, which accounts are more likely to roll forward to higher buckets.
  • Identifying those customers will help in offering the right customers the right payment programs and increase collections.
  • Development included target variable creation by roll rate , variable selection and hyperparameter tuning.
  • Validation was done in both in time and out of time samples.

Associate II

FICO
Bengaluru
05.2019 - 12.2019
  • Responsible for end-to-end development of Credit Risk models for Australian Energy Client to optimize the collections process.

Analytics Consultant

Bridgei2i Analytics Pvt Ltd
Bangalore
11.2016 - 05.2019
  • Responsible for end-to-end development of Credit Risk and Marketing models, used for optimization of Collections and Marketing Strategies.
  • Worked closely with business stakeholders to ensure timely delivery of models serving business purpose.
  • Developed a Logistic Regression Model to identify the customers who will pay back digitally on their own after bouncing in the PLCS portfolio of one of India's Largest NBFCs.
  • This will lead to better treatment allocation on the bounce pool, resulting on decrease in collections cost with same collections efficiency.
  • Creating homogeneous segments for the existing consumer durable customer base for an India based NBFC in order to enhance better understanding of different kind of customer behavior.
  • The idea was to target the right customers for cross sell and upsell while providing maximum value to customers and increasing revenue without the recurring cost of many marketing channels.
  • The whole exercise included selection of suitable representative sample, variable selection by oblique principal component analysis, clustering exercise through K- Prototype method, profiling and validation of the clusters.

Business Analyst

Genpact
Bengaluru
09.2015 - 10.2016
  • Responsible for end-to-end development and monitoring of Credit Risk models.

Education

Master of Science - Statistics

University of Calcutta
Kolkata
06-2015

Skills

  • Credit Risk Model Development
  • Statistical Modeling
  • Machine learning
  • Collections scorecards
  • Account management scorecards
  • Propensity models
  • Customer segmentation
  • Stakeholder Management
  • SAS
  • Python
  • AWS SageMaker
  • Problem solving
  • Stakeholder collaboration

Timeline

Senior AI/ML Associate

JP Morgan Chase India Pvt Ltd
12.2019 - Current

Associate II

FICO
05.2019 - 12.2019

Analytics Consultant

Bridgei2i Analytics Pvt Ltd
11.2016 - 05.2019

Business Analyst

Genpact
09.2015 - 10.2016

Master of Science - Statistics

University of Calcutta

Similar Profiles

CREATE PROFILE
Soham Das