Juan Diego Castro Miyashiro

Juan Diego Castro Miyashiro has a diverse work experience. In 2021, they worked as an AI/ML Software Development Specialist at Health Gauge, where they designed a Twin Neural Network that achieved state-of-the-art MAE for featureless systolic and diastolic blood pressure predictions using raw PPG and ECG signals. Juan Diego also trained an Object Detection Neural Network for a Covid-19 self-testing report system using Faster RCNN and YOLO, approximated the uncertainty of the neural network’s predictions using the Monte Carlo Dropout Bayesian uncertainty estimation algorithm, managed model versioning, parameter tracking, and code reusability with MLflow, and ran Explainable AI algorithms such as LIME and Integrated Gradients to debug DNN performance. In 2020, they worked as a Machine Learning Intern at Health Gauge, where they developed a custom-design Generative Adversarial Network that allows the controlled synthesis of PPG and ECG signals conditioned on age, weight, and heart rate, as well as developed data processing tools to clean the low-frequency and high-frequency artifacts present in the biosignals from the Health Gauge Phoenix Smart Watch. In 2020, they also worked as a Computer Science Teaching Assistant at the University of Alberta, where they guided the lab sections D11 and D05 explaining and providing assistance to students in solving simple python problems for CMPUT 174, and marked and provided feedback on assignments, quizzes, and midterms. In 2019, they worked as a Physics Summer Research Assistant at the University of Alberta.

Juan Diego Castro Miyashiro attended the University of Alberta from 2016 to 2020, where they earned a Bachelor's degree in Physics and Computing Sciences.