Hi, I'm Christopher
I'm a Software Engineer

WebSockets are hard to scale and easy to drop. Twine solves both. It's an open-source real-time service I co-created with three other engineers across different parts of the world. Twine sits between an application's backend and its users, takes WebSocket traffic off the main servers, scales itself, and makes sure a dropped connection doesn't mean lost data.

That last part matters for apps where every message counts, like a telehealth chat on spotty clinic Wi-Fi. A client can reconnect up to 24 hours later and receive every message it missed, in order. Recent messages come straight from a Redis cache, and anything older than two minutes is recovered from DynamoDB.

We load tested Twine to 96,000+ concurrent virtual users over 20 minutes, and it auto-scaled from 1 to 4 servers to handle the load. Getting there meant solving the classic problems of running Socket.IO across many servers: sticky sessions so WebSocket upgrades land on the right server, and a Redis adapter so a message published to one server reaches clients connected to all of them.

Twine runs in Docker on Elastic Beanstalk and deploys into the developer's own AWS account with the AWS CLI and CloudFormation, so they keep full control of their infrastructure and their data. It fills the gap between Socket.IO, which is open source but leaves you to build everything yourself, and Ably, which is managed but takes that control away.

Technologies used: TypeScript, Node.js, Express, Socket.IO, Redis, Docker, AWS (Elastic Beanstalk, DynamoDB, ElastiCache, CloudFormation, ALB)

LoadUp matches moving and junk removal customers with local service providers across the country. I owned and built several of the systems that customers and revenue run through on its Rails platform.

The biggest was the booking API. I built it end to end in 6 weeks, and it now handles 90%+ of the company's customer traffic on a site that sees roughly 10,000 visits a day. I also connected Kustomer AI to our application through a new API, so thousands of quotes get created automatically instead of by an admin, and launched the platform's first quoting system in Rails and React, letting admins send custom-priced quotes that customers view and pay online.

To speed up search, I implemented Elasticsearch search infrastructure and reindexed tens of millions of records into a dedicated Bonsai cluster, cutting common query times by 70%+. I maintained the Stripe payment infrastructure, HubSpot data sync, and Sidekiq reporting pipelines behind $35M in annual revenue across 152,000+ transactions.

For release reliability, I stabilized CI/CD, introduced Jest and React Testing Library for frontend tests, and added Datadog alerts for failed deployments. I also designed an event-driven pub-sub system to consolidate outgoing communications to customers, service providers, partners, and internal users.

Technologies used: Ruby on Rails, React, PostgreSQL, Elasticsearch, Redis, Sidekiq, Stripe, HubSpot, Kustomer, Datadog, Jest, React Testing Library