In a span of one year and a half, I worked at Egnite Health researching and applying Computer Vision models to automate the detection of heart diseases. Such models were developed in TensorFlow from scratch and trained on Databricks. We explored traditional Convolutional Neural Networks as well as multi-instance learning.

In addition to the ML modeling, I developed internal libraries for data wrangling that were HIPAA-compliant. This was critical to ensure the privacy of the patients during model training and evaluation. I also developed internal tools that helped domain-experts to evaluate the performance of our models.

Some other contributions were survival analyses on tabular data to model the risk of heart disseases.