Mohammad Gharaei Email & Phone Number

Suwanee, Ga, Us

United States

- Built and deployed an authorization-to-settlement model to forecast the number of days it would take for authorizations to settle:> Extracted extensive data from databases. Transformed the raw data by generating thousands of relevant features. Conducted EDA, followed by variable reduction techniques such as correlation analysis and PCA. > Trained and tested three types of models, including a deep neural network, a random forest, and a logistic regression, utilizing storage and computational resources hosted on cloud.> Implemented and fine-tuned the deployment of the champion model, specifically addressing the challenges posed by the large size of payment data and the size and characteristics of the feature set. Optimized the feature generation process to enhance speed and efficiency, enabling real-time scoring of authorizations. Leveraged parallel computing to further expedite the deployment process, significantly improving overall operational performance and enabling timely decision-making. - Created a comprehensive end-to-end model training and testing module in Python, enabling seamless model development for fellow data scientists. The module facilitated various tests such as k-fold cross-validation, learning curve plotting, segment-level testing, out-of-time testing, stability testing, and sensitivity testing, ensuring robust evaluation of machine learning models.- Led a team of data scientists in training, testing, and deploying a two-stage loss forecasting model. The two stages were a transaction-level model to predict the probability of chargeback, and a conditional reject model (ACH reject given a chargeback), respectively. The model scores were utilized in Loss Forecasting calculations to satisfy accounting and regulatory requirements.- Trained and deployed an auth-level fraud model. The model effectively reduced false alerts and enhanced the efficiency of credit analysts. This improvement significantly reduced losses, and optimized operations.

Atlanta

- Led model development and validation efforts and supervised quantitative analysts for Pre-Provision Net Revenue (“PPNR”) and Credit Loss Forecasting models across diverse portfolios, including Residential Mortgage, Home Equity Line of Credit (“HELOC”), Commercial Real Estate (“CRE”), Commercial and Industrial (“C&I”), Retail Deposits, Commercial Deposits, and Auto Loans.- Collaborated on the development of Model Development standards and policies for a large bank to provide comprehensive guidelines for each step of the model lifecycle.- Assisted in the formulation of Model Governance standards for a mid-sized bank, introducing criteria for measuring model risk, assessing risk ratings, and monitoring model changes. Contributed to the establishment of procedures for model inventory management, approval of risk ratings, and control of model changes.

Atlanta, Georgia, United States

Developed and validated predictive models to forecast loss and income for mid-sized and large banks. The model forecasts were fed into the banks exercises for compliance with regulatory requirements including Comprehensive Capital Analysis and Review (“CCAR”), Dodd Frank Act of Stress Testing (“DFAST”), and Current Expected Credit Loss (“CECL”). Specifically:- Created comprehensive datasets by leveraging relational databases and SQL queries.- Processed and transformed raw data, created large pools of features, conducted visualization and analytics to extract meaningful insights.- Built account-level and portfolio-level Probability of Default (“PD”) models using various techniques, including Logistic Regression and Delinquency Transition Matrix.- Developed Loss-Given-Default (“LGD”) and Exposure-at-Default (“EAD”) models using Fractional Logit, Ordinary Least Squares (“OLS”), ARIMA, and ARIMAX models.- Built an end-to-end model development pipeline to perform model training, testing, and selection. The pipeline streamlined training of large sets of models by providing flexibility to control various hyperparameters to train and test the models, and by facilitating selection of candidate models based on user-defined guidelines.- Authored development, validation, and audit reports, effectively communicating results to management.

Atlanta, Ga

As a graduate assistant, I held supplemental instruction sessions and office hours for the graduate level course Stochastic Risk Management Models. The course revolved around modeling and measuring risk in finance and insurance industries using statistical and mathematical techniques.

Tehran, Iran

- Designed and engineered mechanical systems for a prominent large-scale construction project.- Demonstrated strong leadership skills by effectively supervising the execution of designs during the construction phase.- Conducted thorough contractor supervision, ensuring adherence to project specifications and quality standards.- Generated comprehensive reports for management, providing critical updates on project progress and highlighting potential areas for improvement.- Collaborated closely with cross-functional teams, fostering seamless communication and coordination to achieve project objectives.- Utilized technical expertise to resolve engineering challenges and ensure optimal system performance.- Played a pivotal role in driving successful project completion, delivering high-quality results within specified timelines.

Iran University Of Science And Technoloogy

Biomechanics Lab.

Master Of Science (Ms), Mathematical Risk Management

Master Of Science (M.Sc.), Biomechanics

Bachelor Of Science (Bs), Mechanical Engineering

Activities and Societies: - Led the university team in the national competition for the design and manufacturing of an urban electric car.