I chain the blocks.

Pursuing professional and personal growth via doing open-source software work, attending chocolatier school, learning Japanese in Japan, developing my coffee roasting abilities, and traveling.

Worked as a software engineer on HHVM and surrounding concerns: performance tooling/analysis, compiler fixes, and improving production monitoring and debugging.* Created HHVM Continuous Performance Measurement tool and metrics dashboard for testing performance of multiple versions of HHVM in production. The tool identifies causes of CPU and memory regressions prior to code being pushed to production, designed to reduce extremely time-intensive code investigations by engineers.* Contributed to HHVM compiler memory and CPU optimizations and implemented fixes that achieved 0.25-0.5% CPU usage and 0.25-0.5% memory optimizations for serving 3B daily users.* Drove major initiative to reduce HHVM release cycle from >2 weeks to 1 week. Built/improved all release tooling to automate performance and QA testing and collaborated with multiple teams (release engineering, web, open source, data storage and caching) to coordinate improvements and enable teams for accelerated release.* Led feature development of HHVM Automator testing tool, used by multiple teams to run production-level tests for measuring performance of source changes. Built functionality that enabled external teams to leverage the tool and extend it for their own purposes.* Owned HHVM team’s relationship with Servicelab (FB’s sitewide performance testing product and its dedicated team). Collaborated with team to improve Servicelab reliability for HHVM performance testing (e.g. metrics reporting and accuracy) and resolved infrastructure issues to improve experiment completion rates from ~50-60% to ~80-95% for 2000+ daily experiments.

* Established Document Platform Operations R&D team to consolidate department-wide infrastructure efforts, set team processes, and led team’s roadmap planning.* Led multi-region data migration efforts to move all uploaded images into Amazon S3 and built Python solutions to migrate 1 PB of image data with 0% data loss.* Drove multiple telemetry efforts. * Assembled a three-member team to rewrite an unreliable telemetry system into a globally distributedsystem via custom Rust services and modifications to open-source software. * Wrote telemetry system proxy server using Rust to add authentication atop StatsD and Carbon. * Crafted custom telemetry collectors and log processors in Clojure and Python atop AWS Lambda. * Rewrote storage system for Cimpress Uploads platform to support .NET Core and allow easy addition of new storage services (GCP, Azure, S3).* Advised teams on implementing applications on AWS and mentored interns and junior hires.* Wrote new continuous delivery framework using F#, Python, Jenkins, and AWS.

Please note that Leaf Holdings was acquired and absorbed into Heartland Payment Systems.* Hired and led a three-person team with offshore members that executed rebuild of legacy infrastructure environments using AWS, Ansible, Packer, and Python to improve fault tolerance and reliability. Improvements resulted in reliability uplift from one 9 to three 9s in months.* Built Python framework to automate creation of cloud infrastructure, built on AWS CloudFormation and Apache Mesos.* Wrote software updater for Leaf’s Android tablet which improved response times and data speeds for end users, using Python, Flask, and AWS CloudFront.* Led technology planning and choice discussions, and reviewed software designs from other development teams.* Performed capacity planning and usage analysis to control resource usage and infrastructure costs.

Owned all software and hardware related to the uploads system that processed 3M files a day and stored >500 TB of customer image data.* Led development efforts to distribute uploads and migrate 350 TB of customer uploads from data centers to S3.* Wrote Cabinet, a generic storage layer built to replace a legacy storage API, reducing LOC by ~80%.* Built Renderly, an automated image regression testing tool built to replace an internal tool. The tool saved >6 months of SDET and QA time.* Created Uploads Preview Cache, a distributed cache system for images built to reduce network load on storage systems, achieving 80% reduction to average image retrieval latency.* Wrote image analysis feature for the uploads pipeline, implementing several algorithms to analyze all customer images for graphical complexity, bounding boxes, and photographic quality.* Implemented system for analyzing uploads for illegal/illicit imagery.

* Built centralized service in Python to create and relate server resources in AWS EC2, DNS, LDAP, and Puppet.* Implemented features for customer lead tracking system, including OAuth support and UI Unicode support.* Sharded a monolithic (20MM records) MySQL database for the HubSpot Leads team.* Contributed to initial rewrite of HubSpot content platform in Django and Python.

Two of my favorite stories in bug finding and fixes came from this job, ask me if you're curious. People I've told the stories to wonder if I'm a wizard or similar.I wrote code for networked simulation software, across multiple products, including:* Network logging and playback* Computer-generated forces (AKA StarCraft for the military)* Network protocol gateways and adapters (i.e. you speak Japanese, I speak English, and this software translates between us)* 3-D visualizer (fancier StarCraft, but noninteractive)* Super everything-including-the-kitchen-sink Solaris project which included pretty much everything above. But on Solaris.I also had to do stuff I wouldn't wish on others:* Rewrite software to work on IRIX, SPARC Solaris, HP-UX, and IBM AIX with a C++ compiler from 2004* Make a complete 3D application work on Solaris 10, a platform the software never supported. My two favorite stories are from this project.* Modify parts of Boost-C++ and various other libraries for all those platforms mentioned in point 1 above.* Did I mention making software work on IBM AIX and Solaris 10?

Chances are that something (or a dozen things) you use runs software that I've written. I worked on security software, and I was the lead developer of the RSA Appliance, a (financially successful) product that was also the black sheep of the RSA product line. I wrote a lot of Java and C++, with a dash of TCL, Perl, and Python, and I got to experience software hell in the form of Apache Struts, Visual Basic 6, and BEA WebLogic.I also developed my love of making people's lives easier at RSA. The original RSA Appliance set up and upgrade took anywhere from one to infinite hours for customers, because it was so manual and obtuse. I rewrote the entire process and it was doable in minutes for users who could read documentation.