Vishal Heliwal Email and Phone Number

Long & Short Term Portfolio Risk Score for a ~$70Bn credit card portfolio of a leading US bank• Developed 2 Xgboost based models to predict likelihood of accounts ending up in BK/Charge-Off/delinquency in a 6-24 month span• Built GBM based models to compare and determine the more suitable technique in terms of performance gain• Applied new forms of information such as Intra-Cycle data and Bureau Triggers in the variable list for the first time• Made recommendations to mitigate… Show more Long & Short Term Portfolio Risk Score for a ~$70Bn credit card portfolio of a leading US bank• Developed 2 Xgboost based models to predict likelihood of accounts ending up in BK/Charge-Off/delinquency in a 6-24 month span• Built GBM based models to compare and determine the more suitable technique in terms of performance gain• Applied new forms of information such as Intra-Cycle data and Bureau Triggers in the variable list for the first time• Made recommendations to mitigate the effect of seasonality in the target variable on the scorecard• Established an explanation for peculiar bad rate trends of the modeling population using FICO and time on file variables• Suggested changes to the existing model building process to make it more intuitive and improve the turn-around-time• Obtained a 6%-11% improvement over the existing model in terms of the top 5% bad capture rate Show less

Collections Model for a ~80Bn credit card portfolio of a leading US bank• Built a Gradient Boosting model to predict movement of credit card accounts from lower delinquencies to closures & charge-off• Invoked best practices such as F-measure analysis, variable clustering & Information value in the model building process• Experimented with segmentation schemas - product portfolio based & account behavior based to build targeted models• Applied Relative Importance cut-offs to… Show more Collections Model for a ~80Bn credit card portfolio of a leading US bank• Built a Gradient Boosting model to predict movement of credit card accounts from lower delinquencies to closures & charge-off• Invoked best practices such as F-measure analysis, variable clustering & Information value in the model building process• Experimented with segmentation schemas - product portfolio based & account behavior based to build targeted models• Applied Relative Importance cut-offs to reduced model complexity up to 90% by removing variables with very low contribution • Assessed model performance on validation samples using metrics such as KS statistic, GINI, AUC and overfitting• Model can rank order accounts based on level risk; strategy can then align suitable channel & intensity of collection efforts• Generated incremental revenue by maintaining a steady book size of spending accounts & increasing dollar amount collected• Model complexity fairly low as compared to generic machine learning models, incremental benefit still sustained Fraud Model for a ~80Bn credit card portfolio of a leading US bank• Built the first In-House third-party application fraud model on credit cards for one of the leading banks in the US • Utilized coefficient significance, chi-square test, VIF & multi-collinearity to select variables for final logistic regression model• Invoked feature engineering to derive new attributes from raw variables to enable better segregation of fraudulent applications• Explored additional data sources from various vendors specializing in fraud domain to enrich predictor variable pool• Tested Machine Learning techniques as alternate modeling approach to evaluate for gain in performance • Fraud Loss mitigation, reduction in operational expenses & incremental account bookings generate $8-$10 million annually • Model outperforms all existing vendor models across multiple time periods, even during a stress scenario Show less

• Undertook transition & supported the implementation of a collections model for one of the leading banks in the US• Assessed performance of 2 models on a newly acquired credit card portfolio by pulling relevant data & applying model equations• Built an application using the library shiny in R to display analysis results based on user inputs on an interactive web interface • Implemented libraries such as Dygraphs, ggplot & shiny themes to enhance aesthetic appeal & interactivity of… Show more • Undertook transition & supported the implementation of a collections model for one of the leading banks in the US• Assessed performance of 2 models on a newly acquired credit card portfolio by pulling relevant data & applying model equations• Built an application using the library shiny in R to display analysis results based on user inputs on an interactive web interface • Implemented libraries such as Dygraphs, ggplot & shiny themes to enhance aesthetic appeal & interactivity of the application• Automated the process of exhibiting results, enabling a comparative study of results through graphical representation • Awarded Employee of the month for an all-round contribution whilst working on multiple projects Show less

Antaragni is the Annual Cultural Festival of IIT Kanpur • Led a 4-tier team of 1000 volunteers to organize a 4-day festival with 40,000 footfalls• Conducted 11 major events & 55 competitions in 4 days, achieving a feat of “4 Days 4 Pronites”• Identified 50+ vendors & coordinated with 300 staff workers, managing a budget of ₹ 35 Lacs • Realized a sum of ₹ 4 Lacs by exploring options to increase campus accommodation by 20%YoY• Procured internationally acclaimed equipment… Show more Antaragni is the Annual Cultural Festival of IIT Kanpur • Led a 4-tier team of 1000 volunteers to organize a 4-day festival with 40,000 footfalls• Conducted 11 major events & 55 competitions in 4 days, achieving a feat of “4 Days 4 Pronites”• Identified 50+ vendors & coordinated with 300 staff workers, managing a budget of ₹ 35 Lacs • Realized a sum of ₹ 4 Lacs by exploring options to increase campus accommodation by 20%YoY• Procured internationally acclaimed equipment for the first ever on-campus Electronic Dance Music• Allocated a budget of ₹ 1 Lac(increase by 200%YoY) to creative team for theme implementation• Explored new venues to allow simultaneous conduction of events & involve a greater audience• Enhanced local visibility of the festival by establishing associations with malls and multiplexes Show less

Galaxy - Annual Inter- Hall Cultural Competition, IIT Kanpur• Led a 4-member team to organize 42 competitions in a span of 6 days; actively participated by 2000+ students• Invited 20 external judges from all over India to ensure insightful judgement & proper feedback to contestants• Augmented the budget by 40 % to ₹ 1.35 Lacs for execution of the events with befitting logistic requirements • Introduced 10 new events; laid out technical themes to engage students from a… Show more Galaxy - Annual Inter- Hall Cultural Competition, IIT Kanpur• Led a 4-member team to organize 42 competitions in a span of 6 days; actively participated by 2000+ students• Invited 20 external judges from all over India to ensure insightful judgement & proper feedback to contestants• Augmented the budget by 40 % to ₹ 1.35 Lacs for execution of the events with befitting logistic requirements • Introduced 10 new events; laid out technical themes to engage students from a non-cultural background• Organized the first Galaxy in 7 years with no grievance filed regarding the conduction of any of events Show less

Model Development for predicting Short Assay of Crudes using Near Infrared (NIR) Spectroscopy • Structured & Analyzed data for 300 properties, pertaining to each of 96 crude samples, as database for new software model • Established parameter values for property value prediction using topological modeling & regressive techniques • Developed different models to increase number of predicted properties for existing crudes & handling heavy crudes separately• Detected shortcomings… Show more Model Development for predicting Short Assay of Crudes using Near Infrared (NIR) Spectroscopy • Structured & Analyzed data for 300 properties, pertaining to each of 96 crude samples, as database for new software model • Established parameter values for property value prediction using topological modeling & regressive techniques • Developed different models to increase number of predicted properties for existing crudes & handling heavy crudes separately• Detected shortcomings in the software & recommended modifications before conversion of the models to an online platform • Achieved results with a deviation of less than 5% between predicted & experimental values of properties Show less

• Contacted 100+ companies, inviting them on campus to avail internship & full time job offers to students of 4 departments and across 5 different programs• Availed internship offers to all 75 students of the Department of Chemical Engineering• Acted as Link Student between the Department of Chemical Engineering & Students' Placement Office in order to reach out to a higher number of core companies• Enhanced the database of companies to be contacted by adding new ones & updating the… Show more • Contacted 100+ companies, inviting them on campus to avail internship & full time job offers to students of 4 departments and across 5 different programs• Availed internship offers to all 75 students of the Department of Chemical Engineering• Acted as Link Student between the Department of Chemical Engineering & Students' Placement Office in order to reach out to a higher number of core companies• Enhanced the database of companies to be contacted by adding new ones & updating the contact details of the existing ones Show less

• Guided a team of 6 freshmen students, ensuring their proper induction to the program • Reached out to 500+ alumni across the globe, acquainting them with the current on-campus Innovations, developments & achievements• Sought donations from the Alumni in order to fund scholarships, travel grants to faculty & students, research work, infra-structure and other miscellaneous campus activities.