Aaryan Samanta

I use machine learning to study biological signals, and I run two organizations that bring AI education to students who rarely get it.

Class of 2028 at Legend College Preparatory in Cupertino, California. Mentored research at Stanford Medicine and MIT, and summer programs at UCLA and UC Santa Barbara.

Displacementhow far the tissue moves
Velocityhow fast it moves
Accelerationhow quickly that speed changes
The three curves at the center of my Stanford project: how far pelvic-floor tissue moves, how fast it moves, and how quickly that speed changes. Move across them with a pointer, or focus them and use the arrow keys. Schematic, not patient data; values are in arbitrary units scaled to each curve’s peak.

Highlights

Papers
Sole author at IEEE AIAM 2025; first author in the International Journal of High School Research.
USACO
Platinum division. Perfect 1000 scores in Bronze, Silver and Gold.
AMC 8 and 10
Perfect scores in 2024 (25 of 25 and 150 of 150). AIME qualifier.
USAPhO
Gold medalist. Second-highest F=ma score in the U.S. (24 of 25).
USABO
Semifinalist (2026).
Kaggle
Silver medal in Stanford RNA 3D Folding Part 2, 92nd of 1,867 teams (top 4.9%).
AI Ethos
Founded a 501(c)(3) that has reached 75+ students in three languages with AI tutoring.

Research

My work sits where algorithms meet biology and medicine: graph neural networks and evolutionary search for classification, genomics of mitochondrial resilience, ultrasound kinematics for a clinical question, and reinforcement learning for emergency response.

Papers

PublishedNov 2025

Quantum-Inspired Hybrid Genetic Algorithm and Graph Neural Network Ensemble for Multimodal Classification

IEEE 7th International Conference on Artificial Intelligence and Advanced Manufacturing (AIAM), sole author

Combines a quantum-inspired genetic algorithm (qubit-encoded chromosomes updated by quantum-rotation gates) with a graph neural network, joins them with a learned late-fusion layer, and explains predictions with SHAP. In the paper’s experiments it reached F1 0.57, against 0.55 for Random Forest and 0.52 for SVM.

Cite this paper
@inproceedings{samanta2025qihga,
  author    = {Samanta, Aaryan},
  title     = {Quantum-Inspired Hybrid Genetic Algorithm and Graph Neural Network
               Ensemble for Multimodal Classification},
  booktitle = {2025 7th International Conference on Artificial Intelligence and
               Advanced Manufacturing (AIAM)},
  year      = {2025},
  month     = nov,
  publisher = {IEEE},
  url       = {https://ieeexplore.ieee.org/document/11322272}
}
PublishedAug 2026

Natural Genetic Variation in Mitochondrial Health-Regulating Genes in the C. elegans Strains CX11314 and EG4725 Leads to Increased Mitochondrial Resilience

International Journal of High School Research, Vol. 8, Issue 16, first author with Mansi Mandal and Mohammed Yasar Meeran

Capstone of the UC Santa Barbara Summer Research Academies (Track 5: Molecular Clock), presented at the Capstone Seminar on July 24, 2025. Why do two wild worm strains resist mitochondrial dysfunction better than the lab strain N2? I combined variant data for 540 isotypes (CaeNDR), consequence prediction with Ensembl VEP, dN/dS selection tests, and 609 RNA-seq samples across six mitochondrial quality-control genes. A neural-network expression predictor reached a Spearman correlation of 0.959.

Cite this paper
@article{samanta2026celegans,
  author  = {Samanta, Aaryan and Mandal, Mansi and Meeran, Mohammed Yasar},
  title   = {Natural Genetic Variation in Mitochondrial Health-Regulating Genes in
             the C. elegans Strains CX11314 and EG4725 Leads to Increased
             Mitochondrial Resilience},
  journal = {International Journal of High School Research},
  volume  = {8},
  number  = {16},
  year    = {2026},
  doi     = {10.36838/ijhsr816.137}
}

Projects and mentored research

CompletedJan to Jun 2026

Functional imaging of the pelvic floor

Stanford University School of Medicine, Department of Urology. Independent mentored researcher. Mentor: Christos E. Constantinou, Associate Professor of Urology, Emeritus.

I analyzed anonymized pelvic-floor ultrasound from 23 subjects and extracted displacement, velocity and acceleration curves during voluntary contraction (“the Knack”), Valsalva and cough, a question that matters for stress urinary incontinence. Preliminary result: voluntary contraction increases urethral support, and cough produced no significant caudal displacement. Not yet published.

OngoingOct 2026 to present

Mentored research in computational linguistics

MIT Department of Electrical Engineering and Computer Science, independent mentored researcher. Mentor: Prof. Robert Berwick.

Weekly research sessions applying machine learning and data analysis to questions in computational linguistics.

Manuscript in progressSummer 2026

InfernoCommand: fire-relative reinforcement learning for wildfire response

UCLA COSMOS, Cluster 1: Brain-Inspired Computing, student researcher

Co-developed and presented as a team poster. An actor-critic agent (a CNN over a 595 × 316 terrain grid plus an MLP over weather and fleet state) that dispatches firefighting resources. Because actions are defined relative to the fire, it generalizes to ignition points it never trained on. The environment is a real Los Angeles terrain model built from SRTM elevation, OpenStreetMap, NOAA weather, and the January 2025 Palisades Fire. Over training, mean reward rose 52% and buildings lost fell 54%. The repository lists its limitations, including synthetic traffic and a few unsourced constants. Manuscript in progress.

CompletedSummer 2025

Molecular Clock: natural genetic variation in C. elegans

UC Santa Barbara Summer Research Academies, Track 5: Molecular Clock, student researcher

Team study of natural genetic variation in mitochondrial-regulating genes in C. elegans, presented at the Capstone Seminar on July 24, 2025. It became the IJHSR paper above.

Completed2025 to 2026

AI/ML capstone

Legend College Preparatory, AI intern (2025) and DS/ML capstone student (2025 to 2026)

Completed a three-course AI/ML capstone: a CNN pneumonia detector, symptom and disease pattern analysis, and a diabetes-risk model comparison.

All research code and papers are collected in ai-research-publications.

Competitions

Contest problems are where I practice proving that an idea works before I trust it. Algorithms, mathematics, physics and biology all count.

Contest results by year
YearContestResult
2026 USA Computing Olympiad

Promoted to Platinum. Perfect 1000 in Bronze, Silver and Gold, ranked #1 among all pre-college participants in each division.

Problems explained, with C++17 solutions

2026 USA Physics Olympiad

USAPhO Gold Medalist. Second-highest F=ma score in the country (24 of 25), which qualified me for USAPhO.

2024 AMC 8 and 10, AIME

Perfect scores on the AMC 8 (25 of 25) and AMC 10 (150 of 150). AIME qualifier.

2026 USA Biology Olympiad

USABO Semifinalist.

2026 ARML Local

Team WARML I-5 placed 2nd globally and 1st among U.S. teams, among 300 teams and 2,000+ competitors.

Official results

2026 Kaggle: Stanford RNA 3D Folding, Part 2

Silver medal. 92nd of 1,867 teams (top 4.9%). The task: predict the three-dimensional structure of an RNA molecule from its sequence.

Competition page

Building and leading

I start things when I see a gap: students without access to good AI tutoring, high schoolers without a global network, and an open question about how the body signals emotion.

Founded2025 to present

AI Ethos, Inc.

Founder, President and CEO. 501(c)(3) nonprofit.

Built multilingual, privacy-first AI tutoring for low-income, rural, multilingual and neurodiverse students. Reached 75+ students in three languages with an average 22% gain in math scores. Received IRS approval in September 2025 and launched a program for San Jose’s Spanish-speaking community in June 2026.

Founded2025 to present

NextGenAI International

Founder and President

Lead a student-run AI education organization offering bootcamps, mentored projects and exchange. Led Taiwan and Thailand roadshows and signed a university MOU. 100+ participants across three countries.

In development

CogArc

Founder and research lead

A wearable and EEG neurofeedback platform testing whether emotion is encoded in how body signals couple (hypothesis stage).

Teaching and service

Ongoing2024 to present

Cupertino Junior Olympiad Math/CS Team

Co-captain and student coach

Coached 20+ peers across 5+ districts toward AIME qualification and USACO Gold.

Ongoing2024 to present

VA Palo Alto Health Care System

Volunteer team leader

Led a student team that reviewed 25+ patient medical images under staff supervision and developed a standardized review workflow to support clinical image access.

Ongoing

Peer tutoring and STEM outreach

Team lead and mentor, AP Calculus Tutoring and Faria STEM Outreach

Tutored 15+ peers in multivariable calculus and graph theory, and mentored 20+ elementary students in coding.

Outside the classroom

Ongoing2024 to present

Track and field, Monta Vista Matadors

Freshman-year school record in the 100 m (11.89) and Rookie of the Year. I anchored the winning frosh/soph 4×400 m relay at the Los Gatos Top 8 Invitational (3:36.02) and ran on the 4×400 m relay that set an all-time stadium record at the RustBuster Invitational. 2026 season bests: 100 m 11.85, 200 m 24.35, 400 m 54.53.

Section leader2025

Percussion, Monta Vista Concert Band

Elected percussion section leader. Directed rehearsals leading to the first-ever unanimous Superior rating across all Monta Vista ensembles at the CMEA Festival.

Languages

English, Spanish and Bengali

Spoken: English, Spanish (AP level) and Bengali (beginner).

Academics

School
Legend College Preparatory, Cupertino, CA. Class of 2028. Track: Computer Science and Engineering with AI and Technology.
Scores and GPA
SAT 1570 (800 Math, 770 EBRW). GPA 4.92 weighted.
AP
AP Scholar with Distinction: 9 exams, 4.77 average score, with 5s in Calculus BC, Physics C (Mechanics and E&M), Physics 1, Physics 2 and Precalculus.
Recognition
NSHSS National Scholar. College Board National Recognition Program, Outstanding Academic Achievement.
Coursework projects
Three-course AI/ML capstone (2025 to 2026): a CNN that detects pneumonia in chest X-rays, an analysis of symptom and disease patterns in patient data, and a diabetes-risk model that compares logistic regression, random forest and SVM. Code and reports.
Tools
Python (PyTorch, TensorFlow, scikit-learn, NumPy, Pandas), Java, JavaScript, MATLAB, Git, LaTeX, Linux.

Sources

Every number on this page can be checked. Where a public record exists, it is linked here. Where it doesn’t, the official record is available on request.

IEEE AIAM 2025 paper
IEEE Xplore
IJHSR 2026 paper
DOI
All publications
Google Scholar and ORCID
Kaggle silver medal
Competition page and its public leaderboard
Track results
Athletic.net
Organizations
aiethos.org and nextgenai.org
Contest scores, AP, SAT, GPA
USACO, USAPhO, AMC and AIME, USABO and AP score reports, SAT scores and transcript: available on request.

Contact

The best way to reach me is email. I’m glad to talk about research, contest problems, or starting an AI education program.

References are available on request.