01 · now
what i'm currently doing
a snapshot of where my time goes right now. last updated september 2026.
project lead @ salt lab, uiuc ischool
took over the project after a phd researcher handoff. analyzed 369k citation passages, found a 40-point swing in how they get labeled, and now testing how that shifts ai model rankings. currently authoring the preprint for submission to qss or jasist.
jepa group @ mit critical data
fused two cardiac foundation models (echo + ecg) on mimic-iv and debugged a 512+-frame mamba × jepa model. co-authored “loud or silent?”, a framework for finding how multimodal clinical ai fails when one modality goes missing. also in the social media + ai group, where i drafted preprints on social-media redesign and ai flourishing.
velvt, co-founder & ceo
a vector-embedding productivity platform. 50+ on the waitlist, accepted to microsoft for startups (~$200k azure credits + a 1:1 mentor) and nvidia inception.
executive & technical director, minorities in stem
40+ chapters and 300+ volunteers, partnered with microsoft csr and first robotics. co-leading national programs that have reached 300k+ students, placed 400+ in project-based learning, run 5+ workshops and hackathons, and raised $3k+.
northstar
every student deserves a north star. a free opportunity-mapping tool for students without a college counselor: a 20-question assessment maps traits and interests, suggests three pathways (a steady fit, a stretch, and an unconventional pick), and ranks verified stem programs, scholarships, and competitions with transparent scoring.
an independent-learning archive
physics, usaco, amc, usapho, ai/ml, statistics, and linear algebra, documenting what i'm teaching myself so the work doesn't disappear.
former
nov 2025 — sep 2026
google-sponsored research assistant @ u michigan-flint
built two kotlin static analyses for divide-by-zero and index errors, annotated 1k+ benchmark cases, and ran experiments on 100+ scraped repositories. submitted our paper to ieee saner 2027 on sep 26, 2026.
spring & fall '26
lead researcher @ nyas junior academy
led a 6-person eeg/bci team on an ml pipeline that classifies motor imagery for prosthetic and fes control, mentored by a dartmouth phd. ieee bigdata hs submission + preprint.
multiforecast
an open-source premier league monte carlo engine. 1.5k+ people and 25k+ views, and a graduate researcher abroad adapted the model for their own work.
