Full-stack Developer in Carrboro, NC, United States

In collaboration with another senior developer, I built the engine that runs hundreds of medical simulations for training doctors throughout the world. We created an authoring system that allows physicians to write the content and logic for running the simulation. We built a logic engine that was tied into a large drug database, providing the brain behind the simulations.

The medical simulations are international with versions in English, Spanish, Italian, and Portuguese. Authored cases are reviewed by medical professionals with specially formatted outputs. User results are recorded into Google Analytics and analyzed for clients.

Our design was optimized for complex medical situations by defining dynamic fields in a relational database. This simplified the schema and minimized changes.

This medical simulation system has scaled tremendously and continues to assist in training physicians worldwide.

I led the development team to migrate a large system from AWS EC2 servers to the GitLab CI/CD systems, deploying through Docker and Kubernetes to containers. I built the Docker files that optimized caching and minimized build times, training other developers to use the same systems locally. I divided the smaller tasks into individual builds. I assisted other dev teams in preparing for future CI/CD migrations.

We needed a system that would support a React application internationally that was optimized with caching in remote edge servers and running an API on AWS EC2 with hosting all the React content on S3 without servers.

AWS CloudFront can handle both static and dynamic content, but there are issues when deploying content from CRA. I built an open-source deployment system to deploy to S3 and CloudFront which uploaded only what has changed and sent invalidation requests to AWS.

Prior to switching to Kubernetes, I built a system to manage multiple deployments of web applications. The client communicated to servers to obtain all deployments. The servers were updated using restricted RSync so that only the changed files were transferred.

Both the client and server-side programs were built in Python 3.6, using the AWS boto3 libraries. This was used daily by all development team members.

Using unique identifiers, this system can get around the limitations of Google Analytics without the use of Big Query. Automated downloads are limited to just seven columns, but this system stitches the limited downloads together.

In an early attempt to provide people with personalized health risk assessments, I built a custom system to convert natural language questionnaires into compiled front ends. The questionnaires were written by medical professionals (never the developers) and could be easily changed to remove all technical complexities. The results were fed into an expert system, along with specific biomarkers. Then the participants were provided with individualized reports.

I set up and currently maintaining an "Open Data Kit" (ODK) server for recording of remote surveys, part of an HIV project in Guatemala. Due to some shortcoming in their deployment system, I provided changes to the ODK project to enable flexible deployments on enterprise systems.