Gaming Cloud:
Designed and developed an elastic compute infrastructure for distributed deep reinforcement learning. Allowed for shared clusters that could scale up to 500 GPU nodes. Led to reduction in training time by up to 75%.
Drove and facilitated cross-team collaboration between Microsoft Research, Playfab, Gaming studios, Azure ML and Windows to enable novel gaming experiences powered by deep machine learning.
Utilized container technologies to enable rapid iteration and deployment of different machine learning frameworks. Worked in multiple first party game code bases to add command interfaces that allowed interactive agent control. Introduced CI/CD which automated building, publishing of containers and managing test runs directly contributing to the quality of the codebase and agility.
Business Application Group:
Lead a squad of 4 to build targeted capabilities that directly accrue to enterprise customers’ ability to uptake the latest release of Dynamics 365 Retail. Worked closely with the Solution Architects, PMs, partners and directly with customer to design and incrementally build features, and extensibility points in POS SPA (Typescript and Knockout).