Full Stack Development with RoR, Javascript - React - Redux - MaterialUI:

• Integrated Microsoft Teams API with Rakuna’s virtual event and interview scheduling features.

• Fixed UI/UX bugs on various product suites with Bootstrap and React.

• Built a SMS Messaging page with Ruby Twilio API client, Anycable Go websocket, React, and Redux-Saga.

• Implemented mass SMS - Email campaign’s front end, back-end, and background jobs for Sendgrid + Twilio

• Provided technical support and hot-fixes per customers’ and QAs’ demands.

• Migrated AWS S3 uploader from deprecated Paperclip to Carrierwave.

• Refactored RSpec for 100% coverage on critical controllers, models, and services after a developing season.

• Building Microsoft Teams app extension for feature integration with Rakuna’s Interview Scheduling product.

Cloud Infrastructure:

• Used Terraform to deploy a functional server on VPC, S3 and EC2 with IAM roles and instance profiles.

• Set up Nginx with SSL and uWSGI for Flask on EC2 to allow secure connections to API endpoints.

• Added critical testing modules to Wercker CI/CD to ensure more reliable code deployment in the future.

• Deployed hot-fixes to production environment using AWS CodeCommit and AWS CodePipeline

Attempt to tackle the Travelling Salesman Problem (TSP) by combining the Divide and Conquer approach and Concorde solver to achieve lower processing time with some sacrifice in accuracy.

The project tested fixed radii as a means of identifying neighborhoods of points. By searching for the best tour to travel all the neighborhoods and then solve for the shortest tour to traverse each neighborhood, processing time decreases. In terms of accuracy, the method was able to provide tours of ~13% longer than the predetermined optimal route - comparative to the 2-opt edge exchange method. The 2-opt's processing time, however, takes a significantly longer time than the neighborhood tour method.

Even though this heuristic method pales in comparison to the state-of-the-art Lin-Kernighan heuristic, I hope this paper opens another approach to the legendary problem TSP.

In the future, applying a dynamic neighborhood recognition system (not fixed radii for each neighborhood) with a Machine Learning model is a fantastic way of increasing the competitiveness of this method.

• Designed training sessions for club members for Model UN conferences in Spring 2020.

• Organize the club’s logistics and communication among club members.

• Building grounds for organizing Luther Mock Model UN in order to simulate conference experience for club members and expand the outreach of Model UN to Luther students.