Improving boot performance and reliability for PowerPoint Live on Teams by building a new set of stateless microservices that fetch PPTs from the file host and communicate with the renderer APIs.

Supporting microservices for PPT Live in Teams with improvements like a Redis cache layer to speed up file downloads and a service monitor setup to prevent regressions.

Research in the SLP lab at Columbia University under Professor Julia Hirschberg

1. Using acoustic and prosodic features from videos scraped from BitChute and DailyMotion, studied how extremist groups on the internet are able to gain followers for their content. My focus for this research was right-wing extremist group QAnon.

2. Developed a model to segment videos by their background music

2. Applied ML models to improve feature quality and deep learning to discern the relationship between video content and viewer response.

3. Designed an emotion classifier to distinguish pro and anti QAnon videos

1. Increase efficiency of data cleaning at for researchers at FIST working on largest participatory farmer remote sensing insurance project addressing climate risk in develop countries

2. Created a timeseries database of flood damage in Bangladesh from social media sources to improve flood index insurance models developed at the Earth Institute

Inspired by students battling COVID and its repercussions, we created CVC to help return a sense of community, safety and normalcy.