hackNY Fellow '20 + Machine Learning Intern @ Atlassian June 2020 - Sept 2020
I was part of the Electronic Systems Engineering Team. I discovered and fixed a major broken component in Tesla's Autopilot testing process, built scalable and low-latency dashboards to query terabytes of data to detect global autopilot hardware manufacturing issues in real time, and developed a Natural Language Processing Flask service to automatically create/assign Jira tickets on the fly for the team's 10 electronic parts and 70 members.
I was part of both the Machine Learning and Backend Teams. I primarily designed, developed and implemented deep learning models based on BERT, Transformers, CNNs, etc for sentence similarity and metaphor paraphrasing, alternate phrase recommendation and context comprehension to reduce false positives in Espressive’s Virtual Service Agent. I also deployed highly optimized inference versions of these trained models as proxy cloud-hosted services that could be queried and used via the Django backend.
I was one of the first employees at Sike Insights, and worked on both the Backend and Deep Learning teams. I helped implement functionality for OAuth, retrieving and parsing emails, as well as generating compatibility between different users based on their personality insights. I also worked on developing a Python wrapper around AWS DynamoDB to ensure smooth symmetric stream encryption.
I was part of a global team of 18 data scientists. I primarily worked on building complex deep learning models to predict whether a user will install the Starbucks mobile application if they're shown a Starbucks advertisement on a different mobile application. My work helped improve install conversation rate by over 30%.