Junior Data Scientist @ Virologic (Now CareX.ai)

Published:

Location

Menlo Park, California, USA (Remote)

Duration

May 2020 - August 2020

About the Company

ViroLogic is a digital COVID-19 screening app that offers daily screenings for COVID-19 symptoms. It is also a personalized passive montoring tool. By keeping track of symptoms and monitoring vital signs on a daily basis, patients can receive an early warning as well as guidance as to how to act in case if their indicators point to a high probability of being infected.

The company has now rebranded and is called CareX.

My Role

I worked with a team of professors, doctors, and graduate students on developing efficient and accurate techniques to measure vitals such as breathing rate, heart rate, SpO2 (oxygen saturation), etc, using only a smartphone’s camera and microphone. Our team was very diverse, containing members from many top universities including Stanford, ETH Zurich, and the Technical University of Munich.

I have managed to develop a breath rate measurement algorithm that gave state of the art results on available public datasets. I have also helped diagnose issues with and drastically improve smartphone camera footage quality to improve performance of all our current techniques.

What I Learned

This was a great learning experience for me since I worked with great mentors and spent a lot of one-on-one time tackling problems together. I have also got to meet many prominent people in the field of computer science including Tom Gruber, one of the co-founders of Siri and Newton Howard, former head of the MIT Mind Machine Project and current director of the Oxford Computational Neuroscience Laboratory.

I have learned a lot from my time at Virologic, including but not limited to:

  • How to more efficiently research read and digest research papers
  • Advanced signal processing methods
  • Eulerian and wavelet transform based video magnification
  • State of the art video and signal processing techniques
  • Using different neural network based approaches (VAE, CNN, MLP) on signals and videos