Mar 2023 – Mar 2024 · 1 yr · Bengaluru, Karnataka, India · On-site

• 1. Application Score: Building the core underwriting model for unsecured business loans using bureau, bank-statement & application variables of the customers for Micro, Small & Medium Enterprises (MSME).
• 2. Behavior Score: Building the underwriting model for existing customers for loan top-ups and loan renewals using the bureau, on-book performance, application & application data.
• 3. Application Score for Thin-files: Building the underwriting model for unsecured business loans using bureau & application vars for thin-file (less / missing bank statement) cases.
• 4. Expected Credit Loss: Built a model to predict PD (Probability of Default) for the current exposure to predict ECL provisions for the next year in accordance with IFRS-9 guidelines.
• 5. Personal Loans Underwriter: Built a model to do risk grading for underwriting personal loans using bureau & personal features.
• 6. Personal Loans Amount Strategy: Built amount strategy for personal loans models.

Mar 2021 – Feb 2023 · 1 yr 11 mos · Bengaluru, Karnataka, India

• Vision Models:
• Mask Detector: Detecting masks for all the UC pros; ensuring safety of the customers & partners
• Identity Verification
• Matchmaking Models:
• Predict acceptance by pros for a service
• Credit Risk Models:
• Predicting credit-worthiness of pros
• Forecasting:
• Category level revenue forecasting
• Personalization:
• Customer level optimization for personalized experience using recommendation systems
• Address Matching:
• Household mapping and coverage using Deep String matching

Apr 2017 – Mar 2021 · 3 yrs 11 mos · Bengaluru Area, India

• Customer Experience Models:
• Developing Voice of Customer models to capture customers' sentiments and vertical-aspects
• Developing Intervention models for increasing customer engagement
• Pricing Models:
• Developed a supervised clustering approach to get Cannibalising group of products for price optimisations
• Developed a pricing model for Clearance sale on FK
• Developing RpC (Revenue per Cost) models for price optimisations
• Product Quality Scoring:
• Product quality scoring using quality metrics like ratings, returns, etc.
• Designed a completely automated and scalable ML framework which churns out around 5K models on daily basis
• ML Framework works in a completely automated fashion without any manual intervention; it performs Forward-Backward feature-selection, hyper-parameter tuning using Random-Search & Coordinate-Ascent and key-metrics predictions using RandomForest and XGBoost
• Gave a talk on the same in Analytics Vidhya Datahack-2018 summit (link attached)
• Selection Modelling:
• Maximise Units by optimising Selection-Quantity for different categories
• Built a Random Forest model using Variable Selection and used Perturbation to marginalise the impact of Selection-Quantity on Units
• Won CEO's Award for Customer Excellence, Flipkart Annual Awards 2017 for Selection Modelling
• Lapsers Survey Modelling:
• XGBoost model to extract main reasons causing Customer Disengagement and Impact prediction for the same

Aug 2015 – Sep 2016 · 1 yr 1 mo · Bengaluru Area, India

• Recommendation Engine for customers basis their order history, hybrid model (content based + restaurant collaborative filtering)
• COD Abuse Model to check Cash On Delivery Abuse
• Cross-Selling of items using Association Rule Mining
• Personalization of restaurant listing basis various levers
• Budget based discovery, tagging of restaurants basis expensiveness
• Restaurant Prep time prediction
• Demand shaping, inventory stock prediction problem to curb order cancels/edits

Mar 2013 – Jul 2015 · 2 yrs 4 mos · Bengaluru Area, India

• Working as Data Scientist in Display Advertising team which handles online advertising for Walmart.com through various partnerships and in-house capabilities.
• Revenue Forecasting for Display Advertising
• User Segmentation using browse/order data to predict user behavior/classify users
• Analytical Framework - Display Advertising
• Managing various marketing campaigns through CPC adjustments

Jul 2011 – Feb 2013 · 1 yr 7 mos · Gurgaon, India

• Developed/Implemented Credit Risk Models for AXP Japan, AXP Canada Portfolio
• Designed/Developed a Javascript application which utilizes LinkedIn API for Prospect-Targetting (Acquisitions) and Facebook API for Viral-Marketing
• Designed/Implemented algorithm to convert a continuous financial attribute to discrete financial attribute using binning while preserving maximum Information Value for the new variable
• Analyzed/Evaluated various techniques like Logistic, Binned-Logistic, Neural Networks, Binned-Neural Networks etc. to effectively and efficiently club various variables into a single logical attribute
• Utilized Cross-Binning technique to enhance existing variables used in Credit Risk Scoring

Jul 2010 – Jul 2011 · 1 yr · Chandigarh Area, India