From production AI platforms to optogenetics experiments to international competitions - built to solve hard problems.
Click any card to dive deeper. PDF reports and live links included where available.
Production AI platform that levels the playing field between policyholders and insurance companies. Cofounded and shipping: a diminished-value claim funnel, a body-shop portal, and an SEO knowledge engine, against a $26B/year consumer loss problem.
FAA & Kenya-certified drone pilot operating at ~100m over the Laikipia Plateau. Captured 20+ hours of 4K predation-response footage, now being analyzed with DeepLabCut and SLEAP AI for individual-level movement mapping. Manuscript in prep for The Naturalist with Dr. Craig Marshall & Dr. Isla Duporge.
Built the analysis and review platform behind the lab's full Percept dataset: 202 deep-brain recordings across 87 Parkinson's and dystonia patients, hunting the control signal that adaptive deep brain stimulation should run on.
A live, explorable diagram of the agent system I designed at Novartis: a persistent memory, 126 specialized agents across 16 divisions, and a guard that refuses to let a single fabricated number reach a scientist. Hover to trace it, or run a query and watch it travel the whole path.
Junior paper and senior thesis in Carlos Brody's lab (HHMI). Modeling how prefrontal and hippocampal populations reorganize as mice learn a decision task, from chronic Neuropixels recordings across every stage of learning.
Designed and ran a closed-loop V-maze experiment in NEU350. Demonstrated that optogenetic Gr64f sweet-neuron activation drives spatial place conditioning in freely walking Drosophila - with evidence of goal-directed, outcome-sensitive responding.
EEG study of locus coeruleus-norepinephrine modulation of neural plasticity in traumatic brain injury animal models. 8 subjects, 3-channel recordings, pharmacological interventions analyzed in MATLAB.
Built an end-to-end drug-target evidence platform in 8 weeks, unifying 20 population-scale genetic databases. Now used daily by the DAx team for gene safety screens and more broadly for target discovery and portfolio triage.
Built vision-language model pipelines on Princeton's DELLA supercluster to analyze 4,500 Ethiopian religious paintings. Adapted BLIP-2, CLIP, and CoCa to produce similarity clusters and structured metadata for quantitative art history.
Analyzing ~1.18M hours of naturalistic infant-parent recordings from 15 families. Building deep learning models to predict executive function trajectories from audiovisual behavioral markers in a $50M NIH-funded longitudinal program.
Analyzed 500+ patient microbiome datasets to evaluate FMT delivery routes for recurrent C. difficile infection. Found 43% outcome improvement with lower GI approach. Presented to 100+ senior physicians.
Lead analyst on a 574-response client survey across 14 NJ mobile harm reduction sites with Prof. Heather Howard and Dr. Ben Scher. Built the full R/Python pipeline and surfaced a statistically significant racial disparity in naloxone access (OR = 1.74, p < 0.05) adjusted for housing, arrest history, and fentanyl use.
Two parallel research projects: a 1,500-word paper on monoclonal vs. polyclonal antibody mechanisms (Antibody Society), and a theoretical microgravity experiment design proposing that weightlessness enhances monoclonal antibody production rates.
Designed and pitched a first-in-class diagnostic platform: an HBV cccDNA-conversion reporter assay that measures whether Hepatitis B can complete the DNA repair needed for persistent infection - a capability no commercial assay offers today. Full TAM/SAM/SOM analysis sized the US opportunity at $900M-1.1B.
National competitions, academic distinctions, and athletic achievements.