I'm a software engineer for the Google Cloud HPC Toolkit based in Chicago. The Toolkit is an open source tool that makes it easy to use Terraform, Packer, and Ansible to provision custom HPC/AI/ML infrastructure in Google Cloud following our best practices.
I joined Google in its Professional Services organization, focused on HPC customers. Examples of my work include:
- a pipeline for a Life Sciences tech company that published golden data sets for use in batch and Kubernetes workloads
- a custom autoscaler for GKE that handled high rates of Spot preemption at the customer's extreme scale; a similar solution was later adopted by Google Kubernetes Engine
Before Google, I worked for Univa (acquired by Altair) where I led its Professional Services team. We helped customers adopt Navops Launch to autoscale Univa Grid Engine clusters in the cloud with a focus on life sciences.
At even higher redshift, I managed a team of DevOps engineers at the Center for Gravitation, Cosmology, and Astrophysics supporting the LIGO experiment. LIGO observes black holes and neutron stars through their impact on gravity rather than by light. We were responsible for delivering computing services critical to the collaboration's scientific objectives. Examples include:
- 6000 core computing cluster running HTCondor
- LIGO GitLab: managing our analysis libraries through proper continuous integration techniques. Over 1000 users migrated to this service, replacing a long list of duplicative (and older) services such as CVS, SVN, Jenkins.
- GraceDB. GraceDB is a web portal (on AWS) that coordinates near-real-time observations by partner observatories when the LIGO/Virgo network detects a gravitational wave event. In 2017, the LIGO/Virgo detection of a binary neutron star meger -- and follow-up by 50 observatories! -- was Science Magazine's Breakthrough of the Year.