New opportunity

Stashfin - Data Scientist/Lead - Credit Risk Modeling

Stashfin · Gurgaon/Gurugram, India

About this role

Stashfin seeks a Lead Data Science & Growth Analytics professional in Gurgaon/Gurugram to connect credit risk, product, and growth analytics for its digital lending and payments products. The role focuses on improving customer acquisition and conversion while optimizing portfolio profitability. The posting specifies 7–15 years of experience; its structured job metadata lists a bachelor's degree.

Responsibilities

Lead acquisition scorecards, behavioral models, risk segmentation, and PD, LGD, and credit cost modeling. Partner with Risk to balance approval rates against GCL, and manage champion-challenger testing and model recalibration.
Develop propensity, intent, and customer lifetime value models; assess CAC against LTV; improve marketing ROI, cross-sell, and personalization through Next Best Action frameworks.
Analyze the customer journey from acquisition and KYC through approval, disbursement, and retention. Address conversion bottlenecks, run A/B tests for pricing, UX, and credit strategies, and improve repeat usage.
Apply machine learning, including XGBoost and LightGBM, to risk and growth problems. Use LLMs and NLP for customer insights, document parsing, and automation; engineer features from bureau, banking, and behavioral data and explore alternative data.
Work with large datasets in PySpark and distributed systems; build scalable pipelines, deploy batch and real-time models, and monitor performance, drift, and stability with technology teams. Mentor data scientists and analysts and collaborate with Risk, Product, Marketing, and Business teams.

Required skills include Python, SQL, PySpark, credit risk modeling, growth analytics and funnel metrics, ML modeling, exposure to LLM/NLP applications, and Power BI or Tableau. Preferred qualifications include fintech, NBFC, banking, or retail lending experience; familiarity with bureau and alternative data; end-to-end model lifecycle experience; and strong business judgment and problem-solving. Success measures include disbursement and AUM growth, approval rates, GCL and credit cost, CAC versus LTV, funnel conversion, cross-sell, and retention.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

Data SciencePythonSQLPySparkMachine LearningXGBoostLightGBMCredit Risk ModelingAcquisition ScorecardsProbability of Default ModelingLoss Given Default ModelingRisk SegmentationGrowth AnalyticsCustomer Lifetime Value ModelingPropensity ModelingA/B TestingFeature EngineeringLarge Language ModelsNatural Language ProcessingPower BITableauMLOpsModel MonitoringDistributed SystemsTeam Leadership

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Balance approvals with credit losses
The role leads acquisition scorecards, behavioral models, risk segmentation, and PD, LGD, and credit cost modeling. The posting specifically calls for partnering with Risk on approval rates versus GCL, using champion-challenger testing and recalibration.
Connect acquisition spend to customer value
Build propensity, intent, and LTV models; assess CAC against LTV; and improve marketing ROI, cross-sell, and personalization through Next Best Action frameworks.
Improve the full customer journey
Analyze conversion from acquisition and KYC through approval, disbursement, and retention. The role includes addressing bottlenecks, testing pricing, UX, and credit strategies, and improving repeat usage across lending and payments.
Take models from data to monitored deployment
Engineer features from bureau, banking, and behavioral data; work with large datasets in PySpark; and collaborate with technology teams on scalable batch and real-time deployment, drift, and stability monitoring. The posting also names ML and LLM/NLP applications.
Lead decisions across teams
Translate business problems into analytical work with Risk, Product, Marketing, and Business teams, while leading and mentoring data scientists and analysts. The stated KPIs span growth, credit cost, conversion, cross-sell, and retention.

Likely challenges to think through

Growth versus portfolio quality
Preparation hypothesis: raising approvals or disbursements may conflict with GCL and credit cost targets. Prepare a way to explain segment-level trade-offs, decision thresholds, and how champion-challenger results would inform recalibration without treating growth alone as success.
Measuring funnel changes without misleading conclusions
Preparation hypothesis: pricing, UX, and credit-strategy tests could affect conversion, customer mix, and downstream losses differently. Prepare to discuss experiment guardrails and how you would assess both near-term funnel movement and longer-term CAC versus LTV or credit outcomes.
Keeping deployed models useful as data changes
Preparation hypothesis: combining bureau, banking, and behavioral features across batch and real-time systems can make consistency and drift difficult to manage. Prepare to describe feature checks, performance monitoring, and when you would investigate or recalibrate a model.

Practice your answers

Technical practice

How would you develop acquisition scorecards and PD, LGD, and credit cost models?

What to coverThe posting calls for acquisition scorecards, behavioral models, and PD, LGD, and credit cost modeling. Suggested preparation: explain target definitions, validation, calibration, and how the models inform lending decisions.

How would you balance approval rates against GCL across risk segments?

What to coverThe role explicitly pairs approval-rate optimization with GCL in partnership with Risk. Suggested preparation: discuss segment-level trade-offs and how you would evaluate approval, loss, and disbursement outcomes together.

When would you replace or recalibrate a credit model?

What to coverChampion-challenger testing, recalibration, and monitoring performance, drift, and stability are stated responsibilities. Suggested preparation: distinguish a data shift from deteriorating model outcomes, then describe evidence for recalibrating or promoting a challenger.

How would you evaluate acquisition using propensity, LTV, and CAC?

What to coverThe posting links propensity and LTV models to customer acquisition, marketing ROI, and CAC versus LTV. Suggested preparation: explain how you would define comparable customer groups and weigh acquisition volume against expected value and credit cost.

How would you design a Next Best Action approach for cross-sell and retention?

What to coverNext Best Action personalization, cross-sell, repeat usage, and retention are named in the role. Suggested preparation: discuss selecting actions from propensity or intent signals while evaluating customer response and portfolio outcomes.

How would you diagnose a drop between KYC and disbursement?

What to coverThe posting covers the journey from acquisition and KYC through approval, disbursement, and retention, and requires SQL and visualization skills. Suggested preparation: describe a stage-by-stage funnel analysis, checks for data or definition changes, and how you would present the bottleneck.

How would you test a pricing, UX, or credit-strategy change?

What to coverA/B testing across pricing, UX, and credit strategies is an explicit responsibility. Suggested preparation: explain the comparison, primary conversion measure, and guardrails for approval, credit cost, or retention before interpreting results.

How would you build and assess features for risk and growth models?

What to coverThe role names bureau, banking, and behavioral data, alternative data exploration, and XGBoost or LightGBM. Suggested preparation: address feature availability at decision time, leakage, data quality, and the trade-off between predictive value and operational complexity.

How would you evaluate an LLM or NLP application for document parsing?

What to coverCustomer insights, document parsing, and automation are the stated LLM/NLP applications. Suggested preparation: describe a measurable task, error review, human oversight where appropriate, and the trade-off between automation and reliability.

How would you take a model from large-scale data processing into batch or real-time use?

What to coverThe posting specifies PySpark, scalable pipelines, batch and real-time deployment, and collaboration with technology teams. Suggested preparation: explain how data and prediction requirements affect pipeline design, and how you would check production performance and stability.

Behavioural practice

Tell me about a time you aligned risk and growth stakeholders on a difficult trade-off.

What to coverThe role bridges risk, product, and growth and calls for balancing approvals against GCL. Suggested preparation: use a real situation, your actions to clarify competing measures, and the resulting decision or outcome.

Describe a time you mentored an analyst or data scientist through a complex problem.

What to coverLeading and mentoring data scientists and analysts is a stated responsibility. Suggested preparation: give the situation, how you coached rather than simply took over, and what changed as a result.

Give an example of turning an ambiguous business problem into an analytical plan.

What to coverThe posting asks the lead to translate business problems into analytical solutions across Risk, Product, Marketing, and Business. Suggested preparation: explain the initial ambiguity, how you agreed on a measure and approach, and the outcome.

Tell me about a time evidence changed a product or marketing decision.

What to coverFunnel optimization, marketing ROI, and data-driven decision-making feature in the role. Suggested preparation: recount the situation, the analysis or test you used to challenge an assumption, and the decision that followed.

Describe a time you worked with engineers to resolve a model or pipeline issue.

What to coverThe role includes production models, monitoring, and collaboration with technology teams. Suggested preparation: outline the issue, your part in diagnosing and coordinating a response, and the observed result.

Sharpen your knowledge

Topics to prepare

Credit risk and approval trade-offs
The role calls for acquisition scorecards, behavioral models, risk segmentation, PD, LGD, credit cost, and balancing approval rates against GCL. Revision exercise: explain how a change in approval thresholds could affect approvals, expected losses, and portfolio profitability.
Growth and customer economics
The posting emphasizes propensity, intent, LTV, CAC versus LTV, cross-sell, and Next Best Action. Revision exercise: outline how you would decide whether an acquisition or cross-sell strategy improves marketing ROI without overlooking credit cost.
Funnel measurement and experimentation
The stated journey runs from acquisition and KYC through approval, disbursement, and retention, with A/B tests for pricing, UX, and credit strategies. Revision exercise: map stage-by-stage conversion metrics and explain how you would interpret an experiment that raises conversion but worsens GCL.
Model lifecycle and stability
The role includes champion-challenger testing, recalibration, batch and real-time deployment, and monitoring performance, drift, and stability. Revision exercise: sketch decision points for promoting a challenger model and for recalibrating it after performance changes.
Feature sources and LLM/NLP use cases
The posting names bureau, banking, behavioral, and alternative data, plus LLM/NLP applications in customer insights, document parsing, and automation. Revision exercise: compare which stated data sources might inform underwriting versus growth, and identify a way to check the reliability of parsed information. The checking approach is a preparation suggestion, not a stated requirement.
Lead-level analytical decision-making
The role involves mentoring analysts and working with Risk, Product, Marketing, Business, and Tech teams. Revision exercise: structure a short decision brief that presents a proposed model or funnel change, its effect on the posting’s KPIs, and the trade-off each team would need to consider.

Skills to revise

SQL funnel analysis
SQL expertise and funnel optimization are explicit. Suggested exercise: write queries over a hypothetical acquisition-to-disbursement event table to calculate stage conversion by cohort and locate the largest drop-off.
Python credit and growth modeling
Hands-on ML with XGBoost or LightGBM and credit risk modeling are stated requirements. Suggested exercise: using hypothetical data, build a PD model and a propensity model, compare their outputs, and explain how each would support a different business decision.
PySpark feature pipelines
The role requires PySpark and scalable work with large datasets, including bureau, banking, and behavioral features. Suggested exercise: draft a PySpark pipeline that joins hypothetical source records by customer and time, produces model-ready features, and checks for missing or duplicate records.
A/B test analysis
Running tests for pricing, UX, and credit strategies is an advertised responsibility. Suggested exercise: set up a hypothetical credit-strategy test with approval, disbursement, and GCL measures; calculate the group differences and state what result would warrant further review.
Power BI or Tableau reporting
Experience with Power BI or Tableau is required. Suggested exercise: create a mock dashboard showing approval rate, GCL or credit cost, CAC versus LTV, funnel conversion, and retention, with a view that makes trade-offs visible to stakeholders.
LLM/NLP document parsing
Exposure to LLM/NLP use cases is required, and document parsing is explicitly mentioned. Suggested exercise: design a small parsing example using hypothetical lending documents, then review extracted fields against the source text and record errors; this exercise is a preparation suggestion, not a specified project.

Your preparation checklist

Build a credit-risk project walkthrough
Prepare a concise example of a scorecard, PD, LGD, or credit-cost model you have worked on. Show data choices, validation, the approval-rate versus GCL trade-off, and any champion-challenger or recalibration decisions. These are explicit responsibilities in the posting.
Sketch a risk-and-growth design exercise
Using a hypothetical lending portfolio, outline how you would segment applicants, test an approval-policy change, and track approval rates, GCL, credit cost, and disbursement growth. Label assumptions rather than implying access to Stashfin data.
Map a customer-funnel analysis
Prepare a one-page plan covering acquisition, KYC, approval, disbursement, and retention. Identify a possible bottleneck, an A/B test for pricing, UX, or credit strategy, and the conversion and repeat-usage measures you would examine.
Explain a growth-model decision
Prepare an example or proposed approach for propensity, intent, or LTV modeling. Show how you would compare CAC with LTV and evaluate a cross-sell or Next Best Action decision; these are stated areas of work.
Demonstrate the model lifecycle
Select a project walkthrough or design sketch showing Python, SQL, and PySpark feature pipelines; a justified ML model choice; batch or real-time deployment; and monitoring for performance, drift, and stability. The posting calls for this end-to-end scope.
Prepare an LLM or NLP use-case proposal
Outline one bounded application in customer insights, document parsing, or automation, all named in the posting. Explain how you would evaluate its output and where human review might be useful; present this as a proposal, not a stated company process.
Gather leadership evidence and employer questions
Prepare a brief example of mentoring analysts or aligning Risk, Product, Marketing, and Business on a decision. Ask which trade-offs matter most for this role, how its KPIs are measured, and what model deployment and monitoring capabilities already exist.

Availability last checked: 5 Nov 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…