Vlad A.

Participating in designing the overall structure and schema of the ml inference systems; Cleaning and processing large streams of complexly structured transactional data; Leveraged SQL for data querying and extraction from relational databases, and PySpark for processing large datasets in a distributed computing environment; Worked on data drift detection layer and along with the team tested several handling techniques; Integrated SHAP and LIME for model explainability, allowing end-users to understand model predictions with transparency, fostering trust in ML decisions; Developed models using logistic regression and gradient boosting frameworks like XGBoost, ensuring high accuracy in predicting default risks; Design and implement credit scoring models based on customer behavior, income, and transaction history; Employed K-means clustering and decision trees using Python’s Pandas for data manipulation and Scikit-learn for model development; Utilized Python and Scikit-learn for prototyping and TensorFlow for deploying scalable models. Applied ensemble learning techniques and neural networks to enhance fraud detection capabilities. Integrated the models into the banking system using Apache Kafka for real-time data streaming and processing; Visualized segmentation results with Tableau, providing actionable insights for marketing strategies. Used Apache Spark for handling large-scale customer data efficiently; Utilize ensemble methods, anomaly detection techniques, and predictive modeling to enhance the accuracy of fraud detection systems; Apply Natural Language Processing (NLP) techniques to categorize and label transactions based on textual descriptions. Enhance transactional data understanding by extracting meaningful information from transaction narratives; Managed model lifecycles from development to deployment using AWS SageMaker for model training and deployment, and Git for version control of model code and configurations; Implemented MLflow for experiment tracking and model versioning, coupled with Jenkins for automating the CI/CD pipeline, streamlining the deployment of updated models; Created custom monitoring layer to provide maintenance and insights on overall flow; Increased the stability of the system, increased the speed of the system by 5 times due refactoring; Facilitated collaborative code quality improvement sessions using GitHub for code reviews, fostering a culture of continuous learning and improvement among the data science team.

Participated in all software development end-to-end product lifecycle phases; Initiated image annotation process; Cloud automation and deployment of data pre-processing, models and output; Data Management: collection, processing, and labeling data, ensuring a robust dataset for model training; Conducted full model development and training cycle; Created detailed reports in order to visualize the current achievements, issues, risks, etc. for the customer; Developed Web application with document upload and processing capabilities, as well as functionality for visualizing results; Build technical documentation for computer vision systems for end-users to understand how these systems work and how to use them; Communicating with external stakeholders presenting results.