Zelalem Abahana

Work History

VP (Sr. Data Scientist)

Wells Fargo

12.2019 - Current

Developed an advanced regulatory concern detection framework integrating rule-based logic, LSTM, BERT, transformer-based architectures, ensemble models, and large language (LLM) data augmentation & error correction to analyze millions of complaints and identify potential regulatory risks across 24 federal regulations.
Built and optimized machine learning sub-models using logistic regression, XGBoost, Random Forests, LightGBM, and deep learning architectures, achieving significant gains in precision and recall for regulatory compliance monitoring.
Designed and implemented a cutting-edge payment fraud detection model for consumer credit cards, leveraging anomaly detection, time-series analysis, graph neural networks, and autoencoders to identify suspicious transactions and reduce false positives in fraud prevention systems.
Implemented scalable text preprocessing and feature engineering pipelines using SpaCy, TF-IDF, and embedding models, enabling efficient classification of structured and unstructured complaint data.
Conducted rigorous performance optimization, including sensitivity analysis, hyperparameter tuning, benchmarking alternative models, and integrating advanced techniques like attention mechanisms, sequence-to-sequence modeling, and real-time fraud detection algorithms, while ensuring alignment with dynamic regulatory definitions through collaboration with legal and compliance teams.
Deployed and owned the daily production process for escalating regulatory concern flags and payment fraud alerts generated by the models, ensuring timely and accurate identification of potential risks while supporting quality assurance and monitoring frameworks with business-aligned KPIs.
Developed, deployed, and automated Small Business Lending and Credit Card account-level transition models using machine learning for forecasting balance, revenue, and losses. This process involved data preprocessing, feature engineering, and model selection using advanced algorithms. Implemented a full CI/CD pipeline to ensure efficient model development, testing, and deployment across environments. Integrated the models into an estimator system for real-time forecasting, ensuring continuous updates and scalable performance across large datasets.
Led remediation of model risk findings, addressing model implementation and monitoring challenges across retail services, personal loans, and small business loan forecasting. Ensured robust monitoring and performance tracking using MLflow and version control with Git.
Developed econometric and LSTM-based forecasting models to predict checking account growth trends for consumer and small business portfolios. Employed Git for collaborative development and MLflow for logging experiments and managing model lifecycle.
Contributed to the development and optimization of a scalable model development engine in PySpark and Python for large-scale model deployment.
Led the development of deposit models using forecasting, decay analysis, liquidity stress testing, and behavioral modeling. Focused on forecasting deposit balances, interest expense, and fee income, while optimizing deposit pricing and fee structures. Applied econometric and machine learning models to capture deposit behaviors, segment customers, and predict fee sensitivities, ultimately enhancing portfolio profitability.
Engineered and deployed a challenger ML model to detect fraudulent credit card payment risk, integrating Light GBM and NLP frameworks.
Devised and implemented portfolio valuation methodologies, including NPV and Discounted Cash Flow (DCF) analysis, for recovery and resolution planning.

Senior Consultant

Deloitte

05.2015 - 12.2019

Developed and implemented Home Equity credit risk scorecard models (Default and Roll-Rate) for one of the largest U.S. mortgage servicers. Employed a combination of statistical models and machine learning techniques to improve default risk predictions and optimize loss mitigation strategies.
Designed and implemented machine learning models for auto insurance claims severity and fraud detection, combining deep learning and traditional statistical challenger models to improve detection accuracy and streamline claims processes for large insurance providers.
Led balance, recovery, and charge-off modeling for DFAST stress testing in retail banking, ensuring regulatory compliance and enabling banks to effectively manage RWAs, loss forecasts, and provisioning.
Developed econometric forecasting models for revenue components in Global Asset Management, focusing on PPNR/CCAR. Delivered comprehensive models that accurately forecasted revenues across diverse asset classes.
Led model validation for fixed income assets and derivatives algorithmic trading models, focusing on testing the robustness of pricing models (Black-Scholes, Hull-White, Heston), interest rate models (Vasicek, Cox-Ingersoll-Ross (CIR)), and credit risk models (Structural Models, Reduced Form Models) used in pricing and risk management. Validated Monte Carlo simulations, stochastic differential equations (SDEs), and binomial tree models for derivatives pricing and risk metrics such as Greeks, Value at Risk (VaR), Credit VaR, and Expected Shortfall (ES).
Developed, deployed, and automated Small Business Lending and Credit Card account-level transition models using machine learning for forecasting balance, revenue, and losses. This process involved data preprocessing, feature engineering, and model selection using advanced algorithms. Implemented a full CI/CD pipeline to ensure efficient model development, testing, and deployment across environments. Integrated the models into an estimator system for real-time forecasting, ensuring continuous updates and scalable performance across large datasets.
Developed econometric and LSTM-based forecasting models to predict checking account growth trends for consumer and small business portfolios. Employed Git for collaborative development for logging experiments and managing model lifecycle.
Contributed to the development and optimization of a scalable model development engine in PySpark and Python for large-scale model deployment.
Engineered and deployed a challenger ML model to detect fraudulent credit card payment risk, integrating Light GBM and NLP frameworks for scoring.
Devised and implemented portfolio valuation methodologies, including NPV and Discounted Cash Flow (DCF) analysis, for recovery and resolution planning.
Led model validation and governance initiatives for financial econometric models, establishing an effective challenge process and introducing challenger models to strengthen model governance frameworks and enhance the robustness of financial models.
Headed the development of financial modeling engines, augmenting traditional econometric models with machine learning alternatives. Delivered technical demos to senior stakeholders, showcasing innovative solutions for predictive modeling and decision support.
Supported complex client engagements by applying machine learning techniques to optimize risk management processes in both retail banking and insurance, delivering customized solutions that addressed unique client challenges and improved business outcomes.

Sr. Financial Analyst

Discover Financial Services

03.2014 - 05.2015

Developed econometric forecasting models for Pre-provision Net Revenue (PPNR) components (Sales & Actives, Late Fees, Reserve Hair-cut, Non-accrual Loans, and Merchandise Payment Rate)
Tested, documented and passed all model-risk-management criteria for all econometric models
Performed financial risk stress testing, capital planning and integration of model outputs for CCAR and DFAST processes
Integrated model outputs into balance sheet projections under different macroeconomic risk scenarios for capital planning
Developed hazard & forecasting models for Bank Deposit components
Effectively implemented projects by collaborating with line-of-businesses and model risk management

Statistical Analyst

Walmart

10.2012 - 03.2014

Developed structural equation models (SEM) that enable integrated forecast system, at enterprise level, to forecast sales, units and labor hours and efficiency metrices such as units sold per labor hour
Sourced data from multiple databases for integrated sales forecast systems that forecasts weekly expected labor hour and sales at the store section levels
Effectively communicated across teams solving project problems and facilitating faster executions
Developed and deployed models for anomaly detection systems and measure for post equipment-replacement energy savings
Played a key part in the development and deployment of energy cost analytics and energy global data-mart

Consultant

GIZ

09.2010 - 09.2012

Led the development of database for sustainable land management program on GIZ’s effort in developing major watersheds
Applied remotely sensed satellite data for land cover change monitoring & evaluation techniques to help assess potential KPI’s in drought reoccurrence
Developed spatiotemporal models for physical features of watersheds to assist in land-use and socioeconomic programs
Delivered statistical analysis consultations to team and clients at various ad-hoc requests

Summary

Overview

Work History

VP (Sr. Data Scientist)

Senior Consultant

Sr. Financial Analyst

Statistical Analyst

Consultant

Education

Master of Science - Economics (Econometrics)

Graduate Certificate - AI Engineering

Master of Artificial Intelligence -

Bachelor's - Economics

Skills

Certification

Personal Information

Timeline

Similar Profiles

PATRICIA M. SCHERERPATRICIA M. SCHERER

Andrew S. HollenbachAndrew S. Hollenbach

Laxmikanth ChittampallyLaxmikanth Chittampally

Mayank OberoiMayank Oberoi

Gloria EscobarGloria Escobar