AI Data Analyst II at Mastercard, NY

Led the creation and deployment of AI-driven analytics solutions processing ~2.5 million transactions per day; improved fraud-detection precision from 92% to 97% and cut false positives by 33%. Built scalable Python (Pandas, NumPy) and SQL data pipelines to ingest and transform high-volume financial data, accelerating insight delivery to risk and compliance teams by 4×. Partnered with product, risk, and engineering teams to build predictive models and NLP tools that auto-approve 80% of merchant onboarding and cut average customer-support handling time by 25%, saving ~2,000 staff-hours per quarter. Trained and validated models using AWS SageMaker and Jupyter Notebooks, deploying them to internal APIs and dashboards for real-time performance tracking. Created interactive Tableau and Power BI visualizations surfacing AI insights and KPIs to senior stakeholders, enabling data-driven strategy shifts worth an estimated $18M in annual fraud-loss avoidance.

Independent Research Graduate at Indiana University Bloomington

Led a predictive modeling project based on Behavioral Economics to forecast stock market patterns with a 29% accuracy improvement. Applied Random Forest (R² = 0.999962) and LSTM (MSE = 1.3599214e-05; Cosine Similarity = 0.999553) to enhance market forecasts, and engineered data processing and analysis in Python (Pandas, NumPy) with ML algorithms such as Random Forest, LSTM, ARIMA, and Decision Tree using TensorFlow, Scikit-learn, and Keras to improve forecasting accuracy.

Machine Learning Developer at COOPVA, TX

Collaborated with data engineers and scientists to create customized product solutions for patients with Alzheimer's disease. Contributed to the 'RemindMe' project through user research and technology development. Utilized an AWS-based ETL pipeline (AWS Glue and EC2) and Ultralytics YOLOv8 for kitchen object detection with data security via AWS S3 and KMS. Transformed and curated 1,000+ images, 20+ Excel files, and medical reports using PySpark for distributed processing, reducing system downtime by 40%. Streamlined data flow with AWS Glue and Lambda for serverless orchestration. Achieved mAP50 of 88.7% and mAP50-95 of 64.9% with transfer learning and data augmentation. Built a YOLONAS-based fall detection model using SageMaker and EC2, enhancing patient safety by 30%. Worked in an Agile environment with Jira.

Data Science Engineer at Defence Research and Development Organization, India

Played a key role in maximizing the Indian Navy's Maritime Traffic Analysis by influencing the Extract, Load, and Transform module, achieving a 64% boost in data processing effectiveness and improved real-time decision-making capabilities. Led the development and integration of Apache NiFi, Kafka, and Hadoop to create a robust structure for video data processing and storage for subsequent ML processing. Leveraged MS SQL Server to process maritime traffic data, delivering a 55% boost in data throughput and a 21% boost to threat detection. Streamlined SQL queries and regulated the import/export of streamed data, reducing operating expenses by 36% and improving data engineering effectiveness.