Developed the process for and responsible for deploying EGP and supporting services in customer cloud environments.

Own the EGP infrastructure (production servers, workflow workers, databases, etc).

Designed and implemented the authorization system for EGP.

Own the model inference pipeline, core data model, and many system abstractions.

Decreased Spellbook oncall incidents by 1000% as stability DRI.

Responsible for all infrastructure and devops.

Architected the Globally Consistent Maps (Github for maps) system, which is the foundation for the large-scale mapping product.

DRI for projection quality, ground height estimation, and data preprocessing, which are all critical technologies for the lidar top-down product.

Proficient with deep learning, 3D geometry, camera physics, reconstruction and localization, data visualization, and traditional computer vision methods.

Familiar with SoTA neural rendering and novel view synthesis techniques (e.g. NeRFs, IBR, etc).

Wrote and graded problem sets.

Mentored teams for final projects.

Held office hours.

Integrated internal machine learning infrastructure with fblearner pipeline to provide more flexibility and customizability for multimodal model training and evaluation.

Replaced Pytorch Ignite pipelines with Pytorch Lightning for greater code simplicity.