Allan Klaus

The project involved implementing a system for the COVID-19 pandemic for PWNHealth, a subsidiary of Everly Health. The company needed to implement all COVID-19-related things as fast as possible to be able to save people's lives. I worked in a team that implemented features such as receiving patients, filtering options, and scheduling appointments for testing or consulting. We also implemented an automated consultation scheduling feature, which replaced the previous manual scheduling done by an admin.

KEY ACTIVITIES
• Maintained the consultation system between physicians and patients.
• Maintained communication with the e-prescription system.
• Scheduled and rescheduled visits.
• Implemented the COVID-19-related system into telehealth during the pandemic.
• Improved integration with DoseSpot.
• Implemented PHIS encryption with cypher.
• Improved sickness forms.
• Migrated all e-prescription services without impacting patients and physicians.
• Implemented the automated consultation scheduling feature.

STACK
• Languages: Ruby and JavaScript
• Databases: MySQL and Redis
• CI/CD: Circle CI
• Docker and Sidekiq

The project involved improvements in a provisioning ecosystem that crashed a lot because of bugs and instabilities.

The system used Resque to provision services in the background, but it did not have a state machine, so when Resque tried to retry, it would always crash. On a typical day, that meant about 50 services, and there needed to be a sysadmin to make fixes on the database manually.

I implemented a state machine on provisioning, so when it fails, it knows where to start with crash-due actions that have already been done. Thanks to this, the number of crashes that needed a manual fix decreased by 90%.

This project involved improvements in the core relations of the email services database.

The database of email services had existed for about 20 years. The modulation was not well-formatted: when features were added, it was very hard for new developers to understand the basic relation. The worst part was that building new mailbox features took about 3 to 4 weeks.

Refactoring could impact all system users: about one million of them per day. Therefore, we created the new structure of the database and migrated users in batches to minimize impact and fix possible bugs caused by these changes.

When the delivery was over, the result reached was:
• All users were migrated without downtime,
• The database became easier to understand, the learning curve of new developers decreased, and onboarding became faster,
• The delivery of new features decreased from 3–4 weeks to one minute, creating two registers in the database.