HW/SW/ML Co-Design for Google Tensor

Building conversational AI for new hardware products. I work on Speech Recognition, Speech Synthesis, and NLP/LLM integrations for the Humane Ai Pin.

* Developed large-scale data ingestion pipelines using Apache Beam and Python for Natural Language Processing (NLP) models used by Google Assistant. Involves building scalable infrastructure to make training neural models easier.

* Built the backend pipeline for serving the Speaker Diarization model in Google Cloud’s speech-

to-text API, allowing users to detect in realtime unique speakers in an audio recording.

* Developed backend speech recognition features in C++ for Google Assistant, such as Language

Identification and Text Normalization, and ensured that these recognizer models ran seamlessly

on Google’s production environment and on devices.

Worked as a Product Manager to support Google Cloud Speech-to-Text (STT) and Text-to-Speech (TTS) APIs. This was a short-term rotation role in which I drove the development of new features in the Cloud UIs for Speech.

To this end, I worked with Engineering, UX Design & Research to propose new features that could benefit our advanced Cloud users in moving from a Speech API to a UI-based approach.

Developed an end-to-end prototype for a diagnostic ophthalmic Virtual Reality headset to

automate common eye exams.

Designed & implemented backend software stack for headset prototype in C#. Wrote VR

applications with Unity, integrated Cloud data pipelines for our devices, and performed analytics/Machine Learning on user eye tracking data.