Amazon
Applied Scientist · Full-time
November 2021 — Present
- Worked on end-to-end research to production of multi-speaker multilingual text-to-speech cloud and edge models
- (2021-22) Released Alexa male voices in more than 10 locales (see this article as an example for it-IT)
- (2022-23) Contributed to the research of high-performance on-device models operating in a cross-lingual fashion and developed more than 60 voice models that are currently deployed in BMW and Mini cars (see this article and this video)
- (2023-24) Led a team of 3 scientists for delivering an LLM-based speech generation solution tailored for the ads use case
- (2024-25) Trained multi-modal LLMs for the release of Alexa+ (see this article) and Amazon Nova Sonic (see this article)
- Published 3 peer-reviewed papers (2 as second author and 1 as first author)
PythonPyTorchAWSTensorFlowMXNet