• Developed and launched a modernized messaging platform to enable seamless communication between buyers and sellers across Android, iOS, and web

• Built and integrated REST APIs to implement key features such as image attachments, unread message count, and bulk message management

• Led the rollout of the web platform, incorporating opt-in/opt-out options, A/B testing, and a phased deployment strategy to ensure smooth user adoption

• Managed high request volumes exceeding several hundred million per day

• Ensured production stability using Grafana for monitoring and setting up alerts

• Mentored interns, overseeing their project from design to production deployment

• Built tools that allow customers to monitor, analyze, and take action against fraudulent activity in real-time using Angular, React, and Node.js

• Created interactive heat maps, graphs, and dashboards to visualize fraud using D3.js

• Led the technical design and development of an integration testing framework based on Selenium that was used by the UI and QA team to write 200+ test cases

• Implemented a framework in Java to query Prometheus and generate snapshots of Grafana dashboards on-demand for secure external display to customers

• Maintained build script, Dockerfile, and Helm config for frontend Node.js server

• Developed an internal dashboard to manage UI releases to client environments

• Ported 10+ internal administrative webpages from AngularJS to Angular 8

• Automated the deployment of UI services using Bash, Groovy, Docker, and Jenkins

• Implemented a baseline LSTM sequence-to-sequence model with attention mechanism for neural story generation using Pytorch (Test Perplexity: 20.97)

• Wrote evaluation metrics in Python to gauge inter- and intra-story repetition

• Worked with PhD researcher to help write, peer-review, and submit research paper

• Paper accepted to AACL 2020 and INLG 2020 (arxiv.org/abs/2010.09935)

• Enforced expected behavior through 20+ bug fixes and 10+ integration tests

• Fixed a widespread performance issue related to lack of parity between frontend and backend object representation

• Fixed a performance issue with dataset version loading related to multiple unnecessary server calls from the frontend