Ryan Aldrich

Led the NLP Engine team that allows Alexa to understand the user’s intention.

• Worked on every aspect of machine learning: from feature engineering, data annotation, and model training, to runtime execution and operational management of hardware and services on a global scale
• Managed 14 software engineers: (9 direct reports, 1 subordinate manager, and 4 indirect reports.) and ~3 interns per year
• Consistently had the organization’s highest engineer satisfaction ratings and lowest turnover rates
• Developed a talented team that earned promotions for 4 sr. engineers, 5 mid-level engineers, and one manager
• Hired talent at all levels from interns to sr. engineers
• Managed underperformance, with only one process leading to termination
• Managed an annual hardware budget of >$20 MM
• Reduced hardware costs by ~10% each year, through successful cost reduction goals
• Worked with internal and external stakeholders (e.g. Disney, Microsoft, licensed software providers, etc.) to set, track, and accomplish product level goals
• Continued doing most things I did as an SDE, with an increased emphasis on delegation, architecture design, and acting as a productivity multiplier

Worked with several different teams to design and implement substantial portions of the NLP engine.

• Automated the machine learning process including: data preparation, training, evaluation, deployment to production systems, automated A/B experimentation against current models with automated promotion to production
• Coordinated multi-team projects through scope management, risk estimation, outcome estimation, and reporting
• Implemented real-time metrics instrumentation and automated emergency mitigation strategies, provided on-call operational support for Alexa services
• Maintained a high quality codebase through extensive design and code review
• Patent No. 10,878,808 - Speech processing dialog management

Worked on a multi-discipline team of engineers, scientists, and linguists to create Alexa's first system capable of understanding conversational context.

• Patent No. 10,446,147 - Contextual voice user interface
• Designed and implemented the procedural rules engine that backs Alexa’s context systems
• Created the team’s framework for automated functional and integration testing
• Built annotation tooling to generate labeled datasets for statistical model training

Developed a system for a voice activated television remote control.

• Built on both the device hardware (C++) and in the cloud (Scala) linked via a GRPC framework (an early version of Google’s Protocol Buffers)

Saved human annotators hundreds of hours by creating a machine learning system to identify patients diagnosed with Stage IV lung cancer, using the free text of clinical notes.

• Presented at the American Medical Informatics Association 2015 Joint Summits on Translational Science