Work

Computer science research-y

(2019-2020) I completed my undergraduate thesis with Prof. Ellie Pavlick, titled “Data augmentation and the role of hardness for feature learning in NLP.” We studied (empirically with synthetic data) the tipping point when models switch from (1) learning simple rules and memorizing exceptions to those rules, to (2) learning more complex and generalizable rules. I received an Undergraduate Thesis and Research Award (UTRA) from Brown to work on this project. We presented an extended poster (here) at the Natural Language, Dialog and Speech (NDS) Symposium at the New York Academy of Sciences. We also submitted this work to EMNLP 2020. We're also working now on another submission on extending this work to natural language and state-of-the-art NLP models.

(2019) I worked with Profs. Ellie Pavlick and Roman Feiman on simulating implicature-like phenomena with neural networks. We studied whether neural networks can predict the alternatives that listeners generate in the Rational Speech Acts (RSA) model. An in-progress repository with these models is here.

(2018-2019) As part of a graduate algorithms seminar with Prof. Paul Valiant, I presented the following papers with classmates: “Probabilistic encryption” by Goldwasser and Micali, “Undirected connectivity in log-space” by Reingold (slides here), “Learnability and the Vapnik-Chervonekis Dimension” by Blumer, et al. and “Neural Tangent Kernel” by Jacot, Gabriel, and Hongler. Miranda Christ and I also wrote summaries of “Hardness vs. randomness” by Nisan and Wigderson (here), Dinur's proof of the PCP theorem (here), and “A Mathematical Theory of Communication” by Shannon (here), and we previously presented “Primes in P” by Agarwal, Kayal, and Saxena.

(2018) Varun Mathur and I built a model (code here) and wrote a short paper (here) for the EmoContext task at SemEval 2019, as part of a graduate course with Prof. Ellie Pavlick.

(2019) A classmate and I completed an analysis (here) of the effects of climate change on afternoon rainfall in the US for a geology course with Prof. Jung-Eun Lee.

(2018) I wrote much of the battery charging logic (here) for EQUiSat, the second-cheapest satellite in history and Rhode Island's first satellite. EQUiSat is currently in orbit and is alive and well.

(2018) I TA'd the first iteration of “Computing Foundations: Data” with Prof. Kathi Fisler. I had the privilege of designing course content, including labs and daily drills.

(2018) I added “Daily Digest” notifications to App Center in an internship at Microsoft.

(2018) I won the Undergraduate Paper Prize from Brown's Center for Contemporary South Asia for a paper about the rapper M.I.A that I wrote for a course with Brian Horton.

(2018) I made a visualization (here) of Washington lumber data for a paper about Kurt Cobain that I wrote for a course with Prof. Bathsheba Demuth.