Linguistics & Biophysics Researcher
Laboratory Learning Program · Dept. of Physics · Dr. Jason Puchalla
Incoming DS freshman at UC Berkeley, Fall 2026.
Co-founder of Segmentum AI. Spun out from research at Princeton Physics with faculty. Reinventing few-shot segmentation.
Research across deep generative models, info-theoretic learning, compbio. Currently into rl and computer-use agents.
Prev. Researcher @ Princeton Physics, NYU, MILA, Stanford's AIMI Center, NJIT, IIT Madras, and funded by US Air Force Research Laboratory.A few results I'm proud of: papers at AAAI, ICLR, ACS, CI CRC, the UN AI for Good Global Summit; Regeneron Science Talent Search Scholar.
I've been playing Carnatic flute for twelve years 🪈
As of June 2026
Princeton Physics Lab — novel few-shot tumor segmentation model work under Dr. Jason Puchalla. Co-founding a startup around the underlying IP. Research filed for IP rights.
NYU & MILA — building normalizing flow datasets with Prof. Ravid Shwartz-Ziv. arXiv preprint out.
NJIT SCIIS Lab — reinforcement learning for robotic navigation and semantic segmentation. Paper accepted at the ASCE Construction Research Congress.
Stevens / AFRL — year two of the Air Force Research Lab's National SDR Challenge. Line-of-Sight MIMO and spatial multiplexing.
Starting at Berkeley in August.
Linguistics & Biophysics Researcher
Laboratory Learning Program · Dept. of Physics · Dr. Jason Puchalla
Deep Learning Researcher
New York Institute · Prof. Ravid Shwartz-Ziv · Info-theoretic deep learning
AI Research Intern
Artificial Intelligence in Medicine & Imaging (AIMI) · Selected from 2,000+ applicants
Reinforcement Learning Researcher
SCIIS Lab · Prof. Xi Hu & Prof. Rayan Assaad
AI & Wireless Communications Researcher
AFRL National SDR Challenge · Signal Seekers Team Captain · $17k USRP grant (×2)
Bioinformatics Research Intern
Dept. of Biotechnology & Cancer Institute Chennai · De novo antibody design
ML Research Intern
Robert Bosch Center for Data Science & AI · 35 DL models for yeast growth prediction
Beyond the Loss Curve: Scaling Laws, Active Learning, and the Limits of Learning from Exact Posteriors
Generative Flow Networks for Lead Optimization in Drug Discovery
An Inclusive Deep-Generative Framework to Overcome Racial & Regional Barriers in Skin Disease Diagnosis
Deep generative flow network driven multi-objective molecular design for accelerating drug discovery
Deep Generative De Novo Design of High-Affinity Antibodies for Therapeutic Targeting of Disease-Associated Antigens
A Framework of RGB-Thermal Sensor Fusion with Zero-shot and Prompt-aware Deep Learning for Enhanced Real-time Thermal Anomaly Detection
ScriptGAN: A Therapeutic Tool for Preserving and Replicating Handwriting Using Generative Adversarial Networks in Neurodegenerative and Learning Disorders
Regeneron Science Talent Search — Scholar
Top 300 nationally. $2,000 scholar award. Nation's most prestigious high school science competition. ↗
American Association for Cancer Research Special Award
Awarded for my research on automated generative antibody discovery.
Regeneron Science Talent Search — Scholar
Top 300 nationally. $2,000 scholar award. Nation's most prestigious high school science competition. ↗
AFRL National SDR Challenge - First HS Student Funded
First and only high school student selected by the Air Force Research Laboratory. Led Stevens "Signal Seekers." $17,000 USRP hardware grant (×2). Selected both years.
Mercer Science & Engineering Fair @ Princeton — 2nd Place
Antibody design research presented at Princeton's CS Department.
NASA Human Exploration Rover Challenge — Top 10 Globally
Team Hercules represented New Jersey at the US Space & Rocket Center, Huntsville, AL. ↗
World AI Competition for Youth (WAICY) — International Finalist
LifeLineAI: smartwatch for real-time heart attack prediction via myocardial EKG analysis.
NASA App Development Challenge — National Finalist
App visualizing the Artemis II mission's real-time flight path and Earth communication links.
Congressional App Challenge Award
BeyondWords AI: reading companion for individuals with dyslexia. Recognized by Congressman Frank Pallone Jr. ↗
Fragments
The most interesting problems don't announce themselves as such.
Research taught me that being wrong quickly is a feature, not a bug.
More here eventually.