Jan 2021 – Apr 2025 · 4 yrs 3 mos · Gurugram, Haryana, India

• ● I lead and own, the Data-Science growth-charter, for following businesses in Paytm
• Currently:
• Lending
• Offline Payments, Merchant Ecosystem
• Insurance
• Device
• Merchant Healthdesk
• Additionally, in past:
• UPI, Wallet
• Recharge & utilities, Rent Payment thru CC
• New User Growth Models
• ● Instrumental in setting-up the team from scratch.

May 2020 – Jan 2021 · 8 mos

Jan 2020 – May 2020 · 4 mos

Jun 2019 – Jan 2020 · 7 mos

Jan 2017 – Jun 2019 · 2 yrs 5 mos

• Generic Risk Score (CIBIL Score) :
• CIBIL Score is treated as a gold standard among generic credit scores in India.
• I led the development of - a new enhanced version (Version-3) of CIBIL-Score (Credit Vision Score) built for Hit-Segment using India's largest credit bureau data.
• The score extensively outperforms all existing offerings in the market and is going to be launched formally very soon.
• Industry based Generic Risk Score for Credit Card, Personal Loan, Two Wheeler, Consumer durable Loan :
• I played variety of roles as a developer, guide, reviewer and ally in development of different Generic Risk Scores which encompass Scorecards for Credit Card, Personal Loan, Two Wheeler Loan and Consumer Durable Loan.

Feb 2014 – Jan 2017 · 2 yrs 11 mos

• New to Credit Risk Score - for Indian Lenders– The objective was to develop a generic risk product that will help lenders to correctly underwrite first time credit seekers who are inherently considered risky.
• # Developed a risk classifier using non-traditional risk-evaluation methodology where applicants with no past credit information were scored using their look alike' information
• # Given financial inclusion initiatives and a very large section of population in India as first time loan applicants, the product helps lenders to avail significant growth opportunity

Jan 2013 – Feb 2014 · 1 yr 1 mo

• New Generic Bureau Product- Built Propensity Scores - for Indian Lenders
• The objective was to develop a suite of generic marketing models that will help lenders to prioritize their marketing efforts and optimize their spends.
• The product gauges a consumer’s likelihood to apply for, or open, a Credit card, Personal Loan or Auto Loan trade
• Models are being developed using India's largest credit bureau data

May 2011 – Dec 2012 · 1 yr 7 mos

• TrueRisk - New P&C product for US Insurance Industry - a Generic Insurance Risk score, to help US P&C Insurance carriers in Underwriting and Pricing decisions
• Instrumental in development of Insurance Score (personal lines) based on consumers' credit behavior using the largest-ever Insurance industry consortium database for US market, created and managed by TransUnion
• Scorecards are developed for both Auto and Home, for various state groups, created according to state based regulations
• Built all Coverage/Peril combined GLM Model on the residual using Credit related variables (on top of the offseted benchmark rate plan) using EMBLEM
• Used GLM technique to model highly skewed, Tweedie distributed claim data
• Product has potential to influence entire personal line business of US P&C industry
• Uses not only historical information about claim experience but also huge credit data (data-pool with ~400 million exposures and a couple of thousand credit/claim experience/policyholder/policy/VIN/performance related variables)
• Technique - GLM (tweedie) on pure premium relativity
• Tool - EMBLEM, SAS

Oct 2009 – May 2011 · 1 yr 7 mos

• Helped the Collection Team of a leading Telecom Service Provider in India to reduce their bad debt from the postpaid customers
• Built Collection Models to identify which customers are not likely to pay on or before the stipulated time and move into worse delinquencies, and also what would be their probable bad debt amount.
• Using the model, designed complete strategy to help the client in maximizing the collection amount (given constrains on collection efforts, budget, customer-interaction, churn-rate).
• The sorecard got rolled out at a national level - First time in Indian Telecom.
• Helped an American Auto Loan company in the recovery of the due amounts from the charged-off account
• Built Recovery Model that would identify well in advance the bad debt accounts and their bad-debt amounts, from all the charged off accounts, using Heckit model with the data which had sample selection bias .

Aug 2006 – Sep 2009 · 3 yrs 1 mo

• Some of the Projects I enjoyed working in:
• Provided customized strategies (Cross-Sell, Up-sale, Retention, Collection, Promotions) for a Bank in Thailand by doing Value Based Segmentation
• Value-Based-Segmentation of Credit Card Customers by slicing the heterogeneous portfolio into homogeneous segments, on the basis of dimensions like Profitability, Risk, Transaction-behavior and Loyalty. Detailed Analysis and Profiling of the segments using the Demographics(Static) and Behavioral(dynamic) information
• Helped a leading Telecom Service Provider in India to take analytics driven decision in the areas of Churn Management.
• Built a logistic regression model to predict which customers are likely to attrite so that retention strategies can be made
• Helped a leading Insurance giant in North America in taking data driven decision in the areas of Reserving and Pricing.
• Built home loss model (Severity and Frequency Model) to estimate the probable claim losses for different perils viz. Fire, Water, and Wind-hail using the Generalized Linear Modeling technique.
• Provided data driven insights to a US Retail giant’s home care business product to optimize revenues and profits
• Built a price elasticity model (Linear Regression Model) to capture impact of prices, distribution and various promotions (Own, Competitor, Variants) on Volume Sales
• Out of the box : DMA Analytics Challenge 09
• The challenge was to develop a scoring algorithm that assigns customers to marketplace segments. Built Multinomial Logit Model which stood seventh on the final leader board