Led application development--set guidelines for reviewing code, mentored junior engineers, made most of the high-level technical decisions. Led the effort to break down a Rails monolith into a separate React frontend, Rails API backend, and multiple supporting microservices.
When we pivoted from an e-commerce site to a video-based social app, I heavily refactored the Rails backend to be performant with large sets of data by bypassing the ActiveRecord models, but still leveraging their scopes, by implementing a framework for defining the json data structures within ruby in a simple way.
Implemented a video transcoding pipeline using AWS Elastic Container Service (Docker), ffmpeg, cython, python and SQS that could transcode the videos into variable bitrate HLS and DASH more quickly than AWS Elastic Transcoder, with better quality, fewer errors on mediastreamvalidator, for less cost, and in a way that was easy to scale.
Implemented an analytics pipeline & recommendation engine using Python, Celery, and Redis that allowed us to use realtime analytics data and an incremental approach to collaborative filtering to update a user's video feed in real time to show them videos they are more likely to keep watching. Using this framework I was able to consistently improve retention for new users by refining the collaborative filtering parameters over time.
Implemented an incredibly high-performance Rust & Redis ORM/API framework for our last pivot. Made extensive use of Rust's metaprogramming tools to create a very straightforward framework to leverage Rust's performance benefits without sacrificing development speed.
Everything I contributed to was made to last, and rarely required any maintenance once implemented. I also ensured that being on pagerduty was stress-free for everyone, while simultaneously having almost complete coverage of any failures--there were rarely any incidents, and they were almost always due to external failures beyond our control.
