Developed core features for the Edison AI Partner Portal which enables

integration of 3rd party AI solutions with GE Healthcare products.

• Led development of UI features to improve customer’s ability to

onboard, view, and manage their AI solutions

• Led development of role-based access for frontend and backend to restrict users’ capabilities based on their roles

Developed core features for the Edison AI Workbench 2.x, a clean slate revision of the 1.x platform for building and deploying AI models.

• Researched and guided team in developing UI components and services using Angular 7.x as well as writing unit tests using Jest

• Implemented services for managing authentication, authorization, and user sessions

• Enabled customers to create annotation workflows whereby radiologists can individually or collaboratively annotate images and SMEs can review annotations for quality

• Enabled our 2D image viewer to adapt its UX based on the user's role

Developed core features for the Edison AI Workbench 1.x, a platform to build and deploy AI models.

• Implemented AngularJS UI components for managing data ingestion; composing and browsing data collections; executing and monitoring experiments for training AI models

• Implemented a Node.js service which ingests medical image meta data and annotations which can be subsequently searched against to compose data collections for AI model training

• Implemented Node.js service which logs and archives data ingestion transactions required for FDA clearance of AI models

• Enabled offline AI model training by implementing a Java service that exports annotations generated in the Workbench

Led the creation of a Selenium integration testing infrastructure for an AngularJS medical imaging framework.

Enabled the sale of GE Healthcare products to the Department of Defense (DoD) by modernizing its enterprise authentication software.

• Implemented a modular password rules framework so that new rules could easily be created to satisfy DoD requirements

• Refactored the password hashing mechanism to use PBKDF2 instead of the legacy MD5 algorithm, thus reducing the success rate of a brute force attack