I'm a student, software developer, and hackathon attendee from Georgia Tech. Here at Tech, I'm pursuing the Intelligence and Theory concentrations. In the future, I look to leverage the skills I gain in Quantitative Finance or Computational Biology. Till then, I look to gain industry and research experience in tech, finance, and bioinformatics.
In my free time, I play ultimate Frisbee, go for runs, and listen to new music. You can always talk to me about philosophy, consumer tech, or the Marvel Cinematic Universe.
I participated in 2 DataOpens hosted by Citadel and Correlation One. Each time, I was joined with 3 other members to clean, parse, analyze 3 GBs of data. The first competition took place in February 2018. We were given data on 6 cities across the United States with data ranging from accidents to funding for buildings. My team worked to use this dataset in conjunction with data pulled from the US Census to create heatmaps of accidents and traffic patterns. With these we analyzed the city of Atlanta and determined an optimal spot to place a new hospital and predicted the effect it might have on traffic. This report won first prize of $20,000 and secured us a spot in the final round in the coming November. The second competition took place July 2018. We were given 2 GBs of airline data. We decided to examine the effect of the interconnectedness of airline on its stability to external factors (e.g weather, fuel prices). To do so, we to the minimum spanning tree of a distance weighted graph for each airline to see coverage. We concluded there was no significant correlation between interconnectedness and stability.
In a 12 hour hackathon, with two friends, I created a convolutional neural network in Google Tensor Flow and Python to categorically sort product images provided by Home Depot with 91% accuracy. My performance in the competition would lead to my eventual interview with Home Depot's technology center and my first internship.
WeLocate was the product of myself and 3 friend's 36 hour caffeine chocolate coding session. WeLocate at its core is meant to help small and medium businesses looking to expand their single shops. In expansion, one of the most important factors and struggles is finding a location. WeLocate is a webapp that attempts to find a data-driven solution to that problem. It prompts the user to choose a starting location on a map, specify a radius to search, and describe their type of store. Using data collected from the Yelp API and the AWS machine learning studio, WeLocate provides a map of optimal places to start a new business. With WeLocate, the challenge facing local businesses to fend off large chain competition becomes a little less daunting.
Every morning of the summer after my junior year, my friend would pick me up from my local park for our joint research. At times, he would be running late and (as any avid Pokemon fan would do) I took that time to play Pokemon Go. The park boasted a whopping 18 pokemon stops. After a few weeks of collecting my morning Pokemon loot, I wondered the optimal circuit. Conveniently that summer, I was researching robotics swarm solutions and I read a paper discussing the use of swarms for a heuristic solution to the Traveling Salesman Problem. I realized that with some alterations, it could give me an optimal circuit too. So I created a distance weighted graph using Google Maps and a timed walking speed. After accounting for Pokemon Stops resetting every 5 minutes, I unleashed my swarm solution and found reasonable success. :)
In order to get my hands wet with Android Development, I worked to create GT Task Manager. The app takes input from users to create and store tasks and adds them to a real time database. Equipped with a fully functional login authentication, users can edit and rearrange tasks across all devices. The app is equipped with an intelligent schedule designer. Given tasks, each with their own deadline and estimated completion time, GT Task Manager will output an optimal schedule for a user defined working time.
About to start! Writeup coming soon.
In the summer of 2018, I worked at BazaarVoice in Austin, a company that handles product reviews and analysis for essentially all major retailers and brands (except Amazon). At BazaarVoice I worked to create image detection models to detect copyright infringement. I used python, Keras, and tensorflow to create two models to detect elements of copyright detection. The first filtered out photos with text overlaid on top with 92% accuracy and the second detected stock photos with 88% accuracy. Together these models automate up 17% of moderation saving the company roughly $50,000 worth of moderation every year. In the latter half of the summer, I worked to create models to rate the quality of photos, to help choose "featured" photos.
During the spring semester of my freshman year, I worked part-time at The Home Depot's technology center. There I worked to create a metric to asses the quality of Home Depot's search auto suggestions. This involved finding correlations between suggestions usage rates and features. One such feature was the diversity of suggestions. To do so, I used Word2Vec and to create an entropy model to quantify the diversity. Other features included characters saved, distance of the most relevant suggestion to actual search, average distance of the suggestions to the actual search. In the future, these analyses will guide the development of a metric to improve current suggestion models and compare to competitors suggestion models.
The spring of 2018, I became a brother of Georgia Tech's Business fraternity - Alpha Kappa Psi. In the organization, I develop my professional and communication skills. I also am a director of the investments team managing a $10,000 fund, leading financial workshops, and recruiting speakers for the fraternity.
I work with masters student in the Quantative Computational Finance program to run the CFC. I personally handle the club account with student government, organizes budgets, and maintain a ledger of voting membership. I also work to spread undergraduate awareness by hosting joint master and undergraduate computational contests
I participate in the Automated Algorithm Designs Vertically Integrated Project at Georgia Tech. I developed genetic algorithm using the Emade platform and python. The algorithms leverage existing learning algorithms to outperform existing optimization methods.