Built an early warning system to identify product safety by scraping social media using NLP. Improved existing model precision by over 30%.Use BERTweet and fine-tuned multi-class classification models to flag tweets indicating harm.

Created an automated search engine leveraging NLP to help sort through a large database of court cases.Delivered a easy-to-use interface that used information extraction and topic synthesis through a LDA model

Analyzed VIIRS satellite data from NOAA and NASA on nighttime lights, and global data on population and economic status to contribute to the production of VIIRS-derived poverty layer that combines the night-lights data with population data to identify areas that are under-lit. We successfully identified a machine learning based solution to produce high-frequency population and poverty estimates on 300 million data points, that can be used in natural disaster relief aid estimation. This is an exploratory research project and was able to present my research at the American Geospatial Union (AGU) Conference in the Fall 2019 meeting.For GEO, I worked with NASA Black Marble Nighttime Lights Daily, Monthly, and Annual data to validate and assess the efficacy of Nighttime Lights for the production of high-frequency population estimates. This included manual validation of measurements for specific cases and the write up of methodologies and results.

Implemented new HTML email templates for the student prospect lists for the school's admissions.Developing student prospect lists of over 20k to target ideal candidates for Columbia Programs. Responsible for creating and fact-checking the email templates, co-working on potential templates and collaborating with the design team in order to produce the desired product.

I co-teach and create the syllabus for a free accelerated class for high school students taught on the weekend on the Columbia University Campus. The goal is to teach and inspire those who have not had access to learning the basics of computer science and programming. The course syllabus is dependent on the interests of the students that semester, ranging from basic web development to data analysis.

Taught SAT prep summer course in Taiwan to a class of 60 for math, English, along with college admissions guidance counseling. Had an adaptive academic syllabus for each unique class to best fit the students.Lead and maintained a class in a foreign country.

Learned how to create an image to retain consumer attention through ad campaigns reaching 200k+ people, for international high fashion brands. Consistently operated in high-stress, public facing environments as an ambassador of the agency.