On-call engineer for PyTorch Edge AIBench, an automated benchmarking service measuring state-of-the-art AI model performance on battery-powered edge devices, including smartphones and XR prototypes.
Rapidly fixed and preempted service issues, ensuring AI teams met critical deadlines. Performed ad-hoc tasks such as flashing stable XR operating systems, detecting device memory leaks, and overseeing failsafe protocols.
Designed and implemented a load balancer to resolve device usage bottlenecks from large batch jobs on limited devices, guaranteeing availability for users during periods of high traffic and enhancing service availability.
Developed CI/CD pipelines using Chef and bash scripts, facilitating automated deployments across Linux, macOS, Android, and iOS systems and significantly improving development efficiency for AIBench and supported teams.
Developed software utilities in Python to automate label creation processes in production lines, reducing the time required at label printing stations by approximately 90%.
Implemented a linear regression model for unit performance grade yields within short notice, predicting that a business deal would have cost EMCORE millions of dollars to fulfill.
Set up automatic laser engraving for precise serial number marking and part tracking by writing C# code to interface with embedded hardware and an extensive relational database, significantly reducing errors associated with manual tasks.
