during my time in the big ideas lab at duke university, i worked to develop key algorithms and biomarker detection of the lab's own infection detection wristwatch. this allowed me to be heavily involved in the development of this product, and allowed for me to learn the importance of cross-functional team management. i analyzed multimodal wristwatch data and developed a new pipeline using linear algebra frameworks to develop an accurate steps detection predictive model and algorithm.

i also worked heavily with the data engineering team to implement these biomarker detection algorithms into the google cloud via data pipelines and clean, functional code, and allowed me to work in detail with the software engineering team to implement our visualizations in the web application, along with optimizing algorithms for lower loading times on the web-app.

click on our poster to watch our final poster presentation!

i interned via the duke impact investing group with ecolytics, a start-up dedicated to helping companies obtain b-corp certification and ensuring corporate sustainability and social responsibility. here, i conducted SEO, product marketing data analytics, and A/B testing, and provided optimization techniques and developed content strategies to improve organic visibility against competitors.

our consulting to ecolytics resulted in a 25% improvement in customer acquisition and 20% increase in conversion rates. this was my first product data science experience, and it was amazing to play such a hands-on role in ecolytics' site product design. being able to play such a proactive role in a start-up's growth and seeing the real-time implementations of our strategies was an immensely exciting experience, and further solidified my interest in leveraging tech for a positive impact.

see our 50-page final business report here.

the summer before freshman year, i worked at the louisiana state university health sciences center as a statistics intern via the undergraduate summer research program. i worked under dr. yu in the biostatistics department and the tobacco control initiative to analyze racial disparities in e-cigarette usage.

to do so, we used the PATH study wave 5 adult cohort data, with 2,500+ variables and 35,000+ cases. this was a very complex dataset that required a large amount of data manipulation and wrangling, as many values were missing, while others weren't in a usable format (i.e. we had to manually create non-hispanic white, non-hispanic black, and non-hispanic other categories as race and hispanic/latino origin were 2 separate questions).

additionally, e-cigarette smoking was an especially popular topic within public health, so we had to find statistical methods that would allow us to extract insights that had not previously been studied. the answer: mediation analysis! we would identify risk factors of e-cigarette usage and analyze the mediating pathway to see which factors may help explain the e-cigarette usage disparity. i presented our approach and models to a school-wide poster symposium.

this novel approach and application culminated into a publication in mdpi stats.

via the beaver-works summer institute cog*works course, i got my first taste of coding and machine learning. this was a rigorous 4-week course in which we explored ml and ai via visual, auditory, and semantic applications. this was one of the most challenging experiences in my academic journey, but so immensely rewarding.

we worked with discrete fourier transforms, spectrograms, and peak-finding; activation functions and convolutional neural networks (cnns); and large language models, rnns, and seq2seq models. the content was rigorous and intensive, along with weekly projects we had to complete to put these concepts into application.

our final capstone project was a sea animal classification tool, where we fine-tuned a vgg-16 cnn model for visual classification and a resnet-18 cnn for audio classification. this was delivered as a pysimple gui allowing users to input a sea animal image or sound, and be able to see the file's sea mammal classification, a description of the animal, the region where it is most commonly found, and a gallery of additional images.

click on the first image to see our presentation!