Modernized image‑segmentation R&D: Re‑engineered the legacy pipeline around Meta’s Segment Anything v2 (SAM2) and Detectron‑based and Grounding Dino models.

Built a scalable training loop in SageMaker: Orchestrated data ingestion from S3 straight into custom loaders, automated balanced train/val splits, and optimized hyper‑parameters to reach state‑of‑the‑art AP (Average Precision) and AR (Average Recall) scores on a bespoke dataset.

Streamlined data & workflow automation: Scripted CVAT‑CLI and CI jobs to batch‑create tasks, export COCO annotations, and version models/artifacts in S3, cutting cycle time from labeling to model update.

Hardened the MLOps stack: Containerized CUDA‑enabled training environments, enforced IAM‑scoped permissions, and authored diagnostic tooling that surfaces GPU/memory bottlenecks, ensuring reproducible, elastic experimentation.

Led the "Audit Corrections" project, resolving critical UI issues by implementing ng-disable to prevent multiple form submissions, resulting in a 45% reduction in duplicate data entries and improved system stability.

Developed a dispute notification system, integrating backend and frontend technologies to notify auditors of disputed audits in real-time, increasing auditor productivity by 30% through streamlined workflows.

Worked on projects dealing with routing incoming payments handling over 2 million transactions per day (APPIN).

Led the development of new enhanced Core API microservices, from identifying inefficiencies in old services to upgrading to Spring 3.

Took over WU’s Bill Pay project, generating over $900M revenue annually. Dealt with bill payments system for external billers.

Additionally stepped up as team Scrum Master, using Agile practices and JIRA K-Boards to guide over 10 people in tracking project goals.