New York, New York, United States
My team owns Alexa’s storage layer, where all data used for machine learning testing and training is housed. The scale we operate on is massive, and requires us to invent new distributed systems approaches to work around AWS limits that we consistently hit. We are building the next generation of Alexa’s storage platform, optimizing for the latest ML techniques being used by hundreds of Alexa's Applied Scientists that are our customers, including online learning and deep learning.
My team owns Alexa’s storage layer, where all data used for machine learning testing and training is housed. The scale we operate on is massive, and requires us to invent new distributed systems approaches to work around AWS limits that we consistently hit. We are building the next generation of Alexa’s storage platform, optimizing for the latest ML techniques being used by hundreds of Alexa's Applied Scientists that are our customers, including online learning and deep learning.
- Full-stack development on audio hardware interface devices including firmware, kernel drivers, features, and web UI, enriching product and end-user experience. Responsibilities included maintaining and enhancing a large legacy codebase, designing use case scenarios, and managing upstream developers in designing software and hardware features of devices. - Identified a need for proper versioning and release management in driver releases; designed and implemented Jenkins/Perforce system to automatically version drivers, allowing actionable and reliable user feedback and internal tracking. - Designed, planned, and implemented complete C++ and Python end-user API libraries for interfacing with MOTU devices, with comprehensive documentation and examples, enabling customers to integrate functionality of devices into their own systems. - Designed, planned, and implemented OS X driver to dynamically translate SMPTE timecode into MIDI timecode, allowing end-users to slave playback of audio software to incoming timecode signals and create fallback systems to ensure reliability of multiple-device performances.