Unicode data file compression: achieving 40-70% reduction over gzip alone
Created a domain-specific compression algorithm for Unicode data files to optimize WebAssembly bundle sizes, resulting in better compression than standard methods.
Software maniac, "crazy-cat-lady", entrepreneur.
Developing Sourcegraph for the past 6+ years.z
Working double-time on a secret ten-year vision to upend the gaming industry.
Created a domain-specific compression algorithm for Unicode data files to optimize WebAssembly bundle sizes, resulting in better compression than standard methods.
Began working on a new critical feature of Sourcegraph to generate API documentation like pkg.go.dev, docs.rs, javadoc.io, etc. based on LSIF code indexing information.
Spent several months reading GLL parsing papers and implementations, which is one of the most complex types of parsing algorithms, and implemented it in Zig (a low level language.)
Explored runtime-composition of parser combinators in Zig as the basis for a new kind of 'omnipotent' regex engine called Zorex, to be used in the Tridex search engine in the future.
Learned the tradeoffs between Bloom filters, Xor filters, Cuckoo filters, Fuse filters, etc. and [implemented the more optimal variants in Zig](https://github.com/hexops/xorfilter).
Entered an IC role at Sourcegraph, focused primarily on developing an MVP backend for Code Insights, a new feature of Sourcegraph.
Began working on Tridex, a 'better' version Google's Zoekt (“Fast trigram based code search”), leading to a multi-month analysis of Postgres performance and behavior: Postgres regex search over 10,000 GitHub repositories (using only a Macbook.)
I lead and developed Sourcegraph's cloud managed instances offering, including red/black deployment infrastructure, Terraform automation, security, backups, processes for creation/cost estimation/upgrades/maintenance, and training sessions.
Led the Distribution team during a management intermission, while co-authoring Sourcegraph's entire monitoring architecture from the ground up, including a Go pipeline to generate out-of-the-box uniform Grafana dashboards & Prometheus service alerts that any site admin with no knowledge of Sourcegraph could reason about.
Eliminated the highly-problematic part of setting up Sourcegraph called the "management console", which was a plague-like issue often locking site admins out of their own instance and presenting a significant deal-blocking barrier for customer onboarding and ARR growth at the time.
I played a founding role in the Distribution team at Sourcegraph, responsible for making Sourcegraph easy to deploy, scale, monitor, and debug.
Working in tandem with Sourcegraph's CTO, Beyang Liu, I acted as one of the few engineers solely responsible for some of Sourcegraph's largest customers, ensuring their success as a dedicated highly skilled engineer - leading to substantial fiscal wins.
src
CLICreated the Sourcegraph CLI, making it easy to perform GraphQL API requests with auth easily & get JSON back fast, get search results in your terminal, etc. It became a core part of Sourcegraph.
Vecty lets you build responsive and dynamic web frontends in Go using WebAssembly, competing with modern web frameworks like React & VueJS.
At Sourcegraph, I developed Appdash - an application tracing system for Go, based on Google's Dapper and Twitter's Zipkin.
In 2015, I joined Sourcegraph to help build better developer tools and improve the lives of developers everywhere.
While attending school on the weekends, I spent 4 years working 100% full-time on Azul3D, a game engine in Go - learning an immense amount about software engineering, game engines, and working with others.
In 2011 the Go r59 release (pre-1.0) happened. I began learning Go nearly full-time, switching from Python/Cython - while attending college online on the weekends.