• Built application to record video for running video-based pulse-rate measurement on a subject’s face.

• Established framework for calculating real-time video quality metrics and running face tracking to ensure consistent capture of high-quality videos.

• Designed open-source implementation of AugMix, a data augmentation method for images and videos, in the PyTorchVideo library.

• Benchmarked and optimized speed and accuracy for video machine learning models on Kinetics 400 dataset.

• Improved efficiency of video dataloading, a major bottleneck in video network training flows.

• Designed, implemented, and tested CLI tool in Golang to facilitate usage of a long-term store for large time-series data.

• Connected monitoring service with Timescale Forge, a Database-as-a-Service.

• Designed testing scripts for concurrent queries to a TimescaleDB database.