Databases and Caching team

• Migrating database server mappings from Consul to a platform database metadata table in MySQL, eliminating dependency on Consul and mitigating upgrade-related issues.

• ​Optimized database performance by developing a periodic job using Golang to defragment whitelisted tables, significantly enhancing data retrieval speed and system efficiency.

• ​Enhanced data management and operational tracking by integrating timestamps into MySQL schema migration tables

across all components, ensuring precise migration timing verification.

• Developed an OCR-based Feature Extraction Model with Microsoft’s LayoutLMV3 to identify and extract service provider names, customer names, addresses and phone numbers from phone bill images, achieving an F1 score of 0.87. Deployed the model on AWS (Step Function).

• Explored paystub misalignments using LayoutLMV3 and developed intelligent heuristics for fraud detection.

COGS 100: Cyborgs and Future under Professor Eric Morgan

• Led the implementation of Label Studio for video annotations, deploying it on-premises using Kubernetes, enhancing the machine learning annotation process.

• Used YOLOv8 from Ultralytics and PyTorch to create Endoscope AI, focusing on detection of middle and inferior turbinates. Achieved a significant improvement in diagnostic accuracy with an F1 score of 0.9 at 85% confidence.

• Developed a Catalog API to seamlessly integrate COCO-formatted datasets with RedCap data, and engineered a custom dataset creation API using Python, FastAPI, and SQLAlchemy to enhance data management capabilities.•

• Cohorted apps for migration from HAMS (older billing system used internally) to Commerce Cloud Platform.

• Integrated Commerce Cloud Platform with Atlassian Marketplace using Kotlin, MongoDB, and AWS DynamoDB.