Data Analyst Credit Risk

Harsha Donda

Email:**************@*****.***

Phone:469-***-****

LinkedIn:www.linkedin.com/in/dondaharsha0625

Professional Summary

Experienced Data Analyst and Fintech Specialistwithover 5 years of success translating complex financial data into actionable insights that drive credit risk optimization, fraud prevention, regulatory compliance, and customer strategy. Adept at leveraging tools like Python, SQL, Spark, SAS, and Hive to analyze large-scale transactional, behavioral, and credit bureau datasets across cloud platforms including Azure, AWS, and Databricks. Proven track record in building data models and KPI dashboards, conducting cohort and uplift analysis, segmenting risk tiers, and supporting regulatory frameworks (e.g., CCAR, IFRS 9, NPL). Strong collaborator with cross-functional teams including Risk, Finance, Collections, and Strategy to influence key business decisions, reduce delinquency, improve approval precision, and boost portfolio performance.

Education

Masters Degree in Information Systems at The University of Texas at Arlington. Experience

Data Analyst-Capital One

July 2023 – Present

Overview:

Led the design and implementation of a real-time, cloud-native data pipeline to power customer personalization, fraud detection, and credit risk modeling at Capital One. The platform ingested, transformed, and analyzed over 10 TB/month of transactional and behavioral data across multiple sources using a hybrid stack (legacy + modern cloud).

● Analyzed over 1 TB of financial transaction data daily by orchestrating data workflows using SSIS, SQL Server, and Python—uncovered processing inefficiencies that led to a 45% reduction in pipeline latency and faster reporting cycles.

● Developed cohort-based models in SQL and Python to track repayment behavior over time, helping risk teams detect early signs of delinquency and adjust strategies—leading to a 17% improvement in early-stage collections.

● Performed uplift modeling and response rate analysis for credit line increase campaigns, identifying high-propensity segments that boosted campaign lift by 22% while managing incremental credit risk exposure.

● Simulated exposure-at-default (EAD) by incorporating borrower behaviors such as prepayment risk and drawdown usage—contributed to more accurate CCAR stress testing and enhanced capital adequacy insights.

● Conducted loss rate and delinquency segmentation by loan vintage, credit tier, and geography, which revealed 11% excess charge-offs in high LTV buckets—findings were used to refine pricing strategy.

● Partnered with data science and product teams to translate raw signals into engineered features in Databricks, enhancing model input quality and reducing training time by 30%.

● Built Python-based scenario models that integrated macroeconomic variables (e.g., interest rate hikes, unemployment shifts) to forecast expected credit losses (ECL) under IFRS 9/CECL compliance frameworks.

● Created and maintained dynamic dashboards in Power BI and Tableau to track net yield, delinquency trends, cross-sell performance, and credit tier migration—enabling faster product and risk team decisions.

● Conducted segmentation analysis of delinquent and churned customers using SQL and R; discovered a pricing-risk mismatch affecting 18.6% of customers, leading to a policy update that generated $320K in added quarterly yield.

● Analyzed credit policy thresholds and adjusted approval cutoffs based on segmented PD, LGD, and EAD curves—resulting in a 24% increase in approval efficiency and an 11.8% drop in portfolio loss rates.

● Automated and documented complex compliance workflows in Alteryx and Power Automate, reducing turnaround time for monthly risk reports by 30% and improving audit-readiness.

● Designed real-time fraud signal dashboards powered by streaming insights from Apache Kafka and Spark Structured Streaming—supported monitoring of 20M+ transactions/month with <5 second latency.

● Created proactive data quality checks using Great Expectations integrated into Airflow pipelines, cutting production validation errors by 60% and improving downstream analysis reliability.

● Worked closely with Finance and Marketing teams to design optimized data warehouse schemas

(star/snowflake) in Snowflake and Redshift—resulting in a ~50% boost in dashboard and ad hoc query performance.

Data Analyst, BI & Strategy Building- Credit Suisse Oct 2018 – Aug 2021

Overview:

The objective of this project was to develop a data-driven early warning system to flag and segment high-risk accounts prone to delinquency,using historical behavioral data,repayment trends,and macroeconomic indicators.The system aimed to support the Strategy team in portfolio optimization, collection prioritization, and risk-adjusted pricing strategies.

● Conducted deep analysis on customer-level data sourced from transactional systems, credit bureau APIs, and alternative data providers to understand risk profiles, thin file behaviors, and approval funnel leakage.

● Explored and staged raw and curated datasets in Azure Data Lake with partitioning by loan origination date and geography to enable time-series and regional analysis of credit trends.

● Engineered and analyzed predictive features such asCreditUtilRatio, PaymentToIncomeRatio, and FirstPaymentDefault, which improved model liftby 22% and were adopted by credit risk and strategy teams.

● Performed vintage analysis and delinquency roll rate modeling across origination channels and risk tiers, uncovering early-stage default patterns that informed updates to credit policy—leading to an 11% reduction in early defaults.

● Built and maintained interactive KPI dashboards in Power BI and Tableau to track approval precision, yield, and exposure-at-default (EAD) by credit band, enabling portfolio managers to make informed, data-driven adjustments.

● Designed A/B test frameworks using Design of Experiments (DOE) to evaluate new pricing structures and approval thresholds, helping optimize conversion and risk-adjusted yield.

● Analyzed customer lifetime value (CLTV) and profitability segmentation by integrating transaction history, behavioral scoring, and cross-sell activity—resulting in a 19% boost in high-value customer retention via targeted campaigns.

● Used SQL and Python to identify early signs of delinquency and segment borrowers into 30/60/90-day risk groups using ensemble modeling (Random Forest + Logistic fallback), driving early intervention strategies for Collections.

● Delivered risk impact analyses under various macroeconomic stress scenarios (e.g., unemployment rate shocks), providing forward-looking guidance to Risk and Capital Planning teams.

● Conducted monthly portfolio performance reviews and delivered concise risk memos to leadership and Compliance, influencing loan approval bands and provisioning levels.

● Collaborated cross-functionally with Credit Risk, Collections, and Legal to ensure risk signals and FDPs (Forward-looking Default Predictors) met both business needs and regulatory expectations.

● Analyzed fraud patterns and credit risk exposure using log data ingested via Hadoop clusters; optimized Hive queries for faster trend analysis and customer segmentation.

● Translated raw event-level data into customer-level risk profiles using modular Pig workflows, enabling refined segmentation for both underwriting and marketing. Skills

ETL Tools Alteryx, SSIS, Power Automate, Apache Airflow, Azure Data Factory, Databricks (Workflows, Delta Lake), Snowflake, AWS Glue Data Analysis SQL, SAS, Python, R, Bash

Visualization Tools Power BI, Tableau, Looker, SSRS Machine Learning Scikit-learn, XGBoost, TensorFlow, PyTorch, Keras, NLP, PCA, SHAP, LIME, OpenCV, Regression, Clustering, Time Series, A/B Testing, DOE Databases SQL Server, MySQL, PostgreSQL, MongoDB (NoSQL), Amazon Redshift, Snowflake, Azure SQL, BigQuery

Libraries Pandas, NumPy, SciPy, Matplotlib, Seaborn, BeautifulSoup, Streamlit, LangChain, OpenAI API, Hugging Face, REST APIs

Methodologies Agile, Scrum, Jira, and Waterfall

Projects

Real-Time Credit Card Fraud Detection System:

Built a real-time streaming pipeline using Kafka and Spark to detect fraudulent transactions. Applied anomaly detection models (Isolation Forest, Autoencoder) for fraud scoring. Integrated AWS Lambda to trigger alerts with sub-5 second latency. NLP-Powered Customer Support Insights Engine

Built an NLP engine using BERT and spaCy to classify and extract insights from support tickets. Automated sentiment analysis and topic detection for complaint trend monitoring. Integrated with Elasticsearch and dashboards for real-time business visibility. Certifications

● AWS Certified Solutions Architect - Associate

● Azure certified Data engineer

● Microsoft Power BI analyst

Contact this candidate