Nirupam Kar

Director of Engineering

Bengaluru, Karnataka, India15 yrs 2 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Over 14 years of experience in Data Science.
Expert in developing predictive models for financial risk.
Proven track record in leading data science teams.

Stackforce AI infers this person is a Data Science expert in Fintech and Ad Tech industries.

Contact

Skills

Core Skills

Artificial Intelligence (ai)PythonCredit Risk

Other Skills

Logistic RegressionLinear RegressionVowpal WabbitFFM LightGBMXGboostLassoRidgeRandom ForestANNGINIKSInformation valueloglosscross validationmse

About

I have an overall experience of more than 14 years in Data Science used in different businesses like Risk Management to Ad tech. I have worked on different projects like click through rate prediction, revenue per click model development, scorecard building for Acquisition, ECM and Collection, PD,EAD,LGD Model development etc.

Experience

15 yrs 2 mos

Total Experience

3 yrs 9 mos

Average Tenure

8 yrs 4 mos

Current Experience

Media.net

6 roles

Director Engineering

Promoted

Jan 2025 – Present · 1 yr 4 mos

Associate Director Engineering

Promoted

Jan 2023 – Dec 2024 · 1 yr 11 mos

Manager - Data Science / Senior Principal Data Scientist

Jan 2022 – Dec 2022 · 11 mos

Artificial Intelligence (AI)

Principal Data Scientist/ Senior Lead Data Scientist

Jan 2021 – Dec 2021 · 11 mos

Leading DS/ML projects across multiple verticals.
Managing and mentoring 20+ Data Scientists and Machine learning engineers.

Artificial Intelligence (AI)

Lead Data Scientist

Promoted

Jul 2018 – Jan 2021 · 2 yrs 6 mos

Leading development of all machine learning models across multiple teams of Media.net namely Bidding team and the Ad team. I manage and mentor teams consisting of around 10 data scientists and engineers at vertical and horizontal to ensure the development and implementation of different models which are being used to solve different business problems.
1.The team developed and implemented a new system for bidding which takes care of different components like visibility of ad slots, keyword click of a visible impression and revenue per keyword click independently. The new system seems to be generating upto 20% more revenue and 30% more profit across different customers.
2.Developed a model which would identify good urls in terms of RPM for exploration purpose
As a larger part of the role, I have been engaged in helping other teams for different problems around Statistics and Data Science. I am a part of hiring data scientists across the company.
Techniques used:
Logistic Regression, Linear Regression, Vowpal Wabbit, FFM LightGBM, XGboost, Lasso, Ridge, Random Forest, ANN, GINI, KS, Information value, logloss,cross validation, mse, mape,mae,Kmeans, KL distance
Programming Language:
Python, PySpark

Artificial Intelligence (AI)Logistic RegressionLinear RegressionVowpal WabbitFFM LightGBMXGboost+16

Senior Data Scientist

Sep 2017 – Jun 2018 · 9 mos

1. Revenue Per Click Model Development using different techniques like Factorization Machine, Regularized Linear Model, Vowpal Wabbit, LightGBM,Deep-learning etc.
2. Click through Rate Model Development to identify ads with more probability of being clicked after a corresponding contextual keyword got clicked.
3. Conversion rate Model Development for certain ad providers
4. Working with the engineers to ensure smooth and correct deployment or the models and set up an automated tracking process to understand the performance of models
5. Hiring for Data Science team
6. Mentoring Data Scientists and Data Engineers
7. Conducting trainings for Statistics and Data Science
Techniques used:
Logistic Regression, Linear Regression, Vowpal Wabbit, FFM LightGBM, XGboost, Lasso, Ridge, Random Forest, ANN, GINI, KS, Information value, logloss,cross validation, mse, mape,mae,Kmeans, KL distance
Programming Language:
Python, PySpark

Artificial Intelligence (AI)Logistic RegressionLinear RegressionVowpal WabbitFFM LightGBMXGboost+16

Citi

2 roles

Manager

Promoted

Jun 2015 – Aug 2017 · 2 yrs 2 mos · Mumbai Area, India

1. Involved in application scorecard redevelopment project for multiple portfolios mainly leveraging logistic regression, ensemble model etc.
2. Developed behavioral scorecard using Machine Learning Techniques (GBM, Random Forest)
3. Guided the development of Collection Scorecards using GBM method
4. Led quarterly and annual validation of custom application scorecards for 4 Cards portfolios and related works
5. Supervised the disparate impact analysis to be completed annually for all the custom scores currently being used
6. Different adhoc works
Techniques used:
Logistic Regression, GBM, Lasso, Ridge, Random Forest, GINI, KS, Information value, logloss,cross validation
Programming Language:
Python, SAS, RevoScaleR, VBA

Logistic RegressionGBMLassoRidgeRandom ForestGINI+9

Assistant Manager

Dec 2013 – May 2015 · 1 yr 5 mos · Mumbai Area, India

1. Responsible for quarterly and annual validation of custom application scorecard for 4 Cards portfolio and related works
2. Involved in building a legal, compliance and Fair lending compliant Disparate Impact analysis framework from the scratch

Aptivaa

2 roles

Assistant Manager

Mar 2013 – Dec 2013 · 9 mos

Consulting works related to AIRB compliant risk parameter estimation and validation

Senior Consultant

Jan 2012 – Mar 2013 · 1 yr 2 mos

Engaged in development and validation of application and behavioral scorecards of both Retail(Auto,Agriculture,Personal Loan,Business loan, SME etc.) and Corporate portfolios(Manufacture,NBFC, Trade & Services,HNWI etc.) for different large banks of US, Middle East,Brazil, India .
Development and validation of LGD & EAD models for corporate portfolios Banks based on US and India.
Research works related to different financial modeling sectors including Bayesian LGD, Judgmental scorecard, Cutoff Score determination, Sector concentration risk, Propensity Score etc.

Genpact llc

Business Analyst

Aug 2010 – Dec 2011 · 1 yr 4 mos · Kolkata

1.Development of Basel II compliant Probability of Default (PD), Exposure at Default (EAD) and Loss given Default (LGD) models retail (mortgage) and commercial (CRE) portfolio for a large bank based of US.
Linear Regression, Logistic Regression and a combination of them have been used to develope the Models. The models have been validated rigorously to prevent over fitting and under fitting. Matrices like Gini, KS, Divergence Index, Logloss, miss classification rate ,Mean squared Error, Mean absolute error have been used to estimate the errors of the Model
2.Subsequently calculations of Through the cycle PD, Minimum capital requirement, Risk weighted Asset have been ensured.