Hi, I'm Rohan Sen

Information Systems and Computer Science student at Carnegie Mellon University (Class of 2030), research intern in health AI at the Feinstein Institutes for Medical Research, and former captain and lead programmer of FTC Team #8721 Mighty Mechanics.

View my portfolio GitHub LinkedIn Email
Rohan Sen

About Me

I’m a student at Carnegie Mellon University pursuing a B.S. in Information Systems with an additional major in Computer Science, and a 2026 graduate of Stuyvesant High School. I’m interested in artificial intelligence, robotics, and software systems with real-world impact.

My work spans medical AI research, autonomous robotics, full-stack web development, and STEM education. I have built a multi-model deep learning ensemble for chest X-ray pathology classification, a custom Java motion-control library for FTC robotics, and web platforms used by hundreds of students.

2×

FTC Worlds Qualifier

6+

Years Programming

250+

Students Taught at PS 117Q

1,000+

Students Served by stuyarista.org

Portfolio

A selection of the projects I'm proudest of, across robotics, medical AI, full-stack web development and systems programming. Each one lists what I built and the stack behind it.

Featured Projects

Larger, long-running projects where I owned major parts of the design and code.

Mighty Motion control pipeline: odometry feeds the localizer; the robot pose and a motion-profiled path feed pure pursuit, which outputs a target pose; x, y and heading PID controllers turn it into robot velocities; inverse kinematics converts those to wheel speeds for per-wheel feedforward motor controllers
Mighty Motion's control pipeline · click to enlarge

Mighty Motion

Lead Programmer · FTC Team #8721 Mighty Mechanics · 2023–2026

A custom Java motion-control library and robot codebase for FIRST Tech Challenge that replaced the open-source libraries our team had relied on. It ran our robot through the 2026 season, when we won the NYC Championship and the FTC Control Award and qualified for the World Championship in Houston.

  • Built our own localization, motion profiling, holonomic pure pursuit, PID pose control and mecanum inverse kinematics, plus tuning OpModes for each controller.
  • Fused goBILDA Pinpoint odometry with Limelight 3A AprilTag poses through parallel 1D Kalman filters, and isolated a Limelight firmware defect that returned NaN rotations for near-head-on poses.
  • Built AutoAim: a turret angle solver plus a quadratic distance-to-RPM regression for the shooter, so aligning a shot became a single action.
  • Cut autonomous overshoot by regressing braking distance against velocity and decelerating early.
Scatter plot from ProfiledPathGenerator: the discretized cubic-spline path through seven waypoints on the 144-inch FTC field, drawn with JFreeChart
Output of the repo's own code for its sample waypoints · click to enlarge

ProfiledPathGenerator

Solo project · FTC autonomous path planning · 2025

Turns a handful of field waypoints into a smooth path the robot can drive in autonomous. Instead of hand-tuning arcs, it fits a piecewise parametric cubic spline through every waypoint by solving a single linear system, then samples it into dense points that a path-following controller can track.

  • Builds the 8n × 8n constraint system directly: waypoint continuity (assembled with a Kronecker product), C¹ and C² continuity at every interior knot, and zero velocity at the start and end.
  • Solves for every spline coefficient at once with EJML, then discretizes each segment into 50 points.
  • Plots the result on the 144-inch field with JFreeChart, so a route can be checked before it runs on the robot.
  • Later ported into Mighty Motion, where I connected it to a geometric path-following controller and live dashboard graphing.

stuyarista.org

Vice President of Web · Stuyvesant ARISTA · 2024–2026

The website and backend for Stuyvesant's National Honor Society chapter, used by 200+ members to log volunteer credit and by 1,000+ students to request peer tutoring. I led the 15-person web team and wrote about 90 commits to the rewrite.

  • Built cancelable tutoring requests, admin deletion of unclaimed requests, and fixes for request spamming on the tutoring page.
  • Wrote the credit-rollover verification script and worked on the credit-calculation logic used to track membership requirements.
  • Rebuilt the interactive club fair map, and added dropdown navigation and loading screens.

Cogni

Co-Founder & CTO · 2024–2025

An AI study platform that turns any text into flashcards and adaptive practice tests. It started as our 2024 Congressional App Challenge (NY-12) submission and grew into a production app at gocogni.io.

  • Prototype: a vanilla HTML/CSS/JS app calling the OpenAI API for flashcard and interactive test generation.
  • Production: a React 19 + Vite frontend and a Hono.js API, both deployed on Cloudflare Workers, with Supabase for Postgres and Google OAuth, and GPT-4o for generation.

More Projects

Team and solo builds from software development, systems and graphics coursework, plus side projects.

NerdyMap showing live subway, LIRR and Metro-North trains across New York City and Long Island, with a panel of nine active service alerts

NerdyMap

NYC transit guide · Project Manager · team of 4

A transit guide for NYC commuters that goes past Google and Apple Maps, showing live train positions, delays, service alerts and elevator/escalator status across the subway, LIRR, Metro-North and NJ Transit.

FlaskGTFS-RealtimeSQLAlchemypandasJavaScript
GitHub
c_chat demo video thumbnail

c_chat

Systems programming · team of 2

A multi-user chat app in C with a TCP server built around select(), switchable group chats, and history saved to disk so users pick up where they left off. I built the GTK4 desktop client.

CSocketsGTK4Linux
Demo video · GitHub
Animated render from the engine: a Phong-shaded sun, an orbiting earth eased with accel, and a moon animated with tween

3D Graphics Engine

Computer graphics · solo

A 3D software renderer in Java, built up stage by stage over a semester: line drawing, transformation matrices, Bézier/Hermite curves, polygon meshes, a coordinate-system stack, scanline-filled solids and lighting. It runs scripts in the MDL animation language, which I extended with Phong shading, tweening and an accel command for eased motion.

JavaLinear algebraParsers
Rendered by the engine from solar_system.mdl · school repo (private)

IKEAlyzer

Data visualization · team of 4

Compares IKEA products across countries by price, currency, rating and review count. It scrapes and cleans about 4,000 product groups into SQLite and renders interactive charts per item.

FlaskSQLiteD3.jsBootstrap
GitHub

GoTravelAI

iOS app · solo

A native iOS trip planner. It turns a destination and preferences into a day-by-day itinerary with Gemini 2.0 Flash, then adds photos of each stop from the Google Places API.

SwiftSwiftUIGemini APIGoogle Places
GitHub

Asteroids3D

Game · team of 2

A 3D take on Asteroids with OBJ model loading, textured meshes, 3D collision detection, sound effects and scoring.

ProcessingJava3D
GitHub

Research Experience

My research focuses on medical AI and computer vision, with hands-on work in deep learning systems for chest X-ray analysis and prior engineering research in materials reuse.

June 2025 - Present

Research Intern | Feinstein Institutes for Medical Research, Northwell Health

Division of Health Artificial Intelligence — Mentors: Dr. Theo Zanos and Dr. Nabil Ettehadi

Project: Multi-model deep learning ensemble for 14-label thoracic pathology classification on NIH ChestX-ray14

  • Built a nine-model PyTorch ensemble on Databricks for 14-label thoracic pathology classification across 112K NIH ChestX-ray14 images.
  • Designed a hybrid CNN–Vision Transformer architecture with residual squeeze-and-excitation stem blocks, mixed-precision training, and a cosine learning-rate schedule with warmup.
  • Built a dual-scale U-Net with Performer attention and EfficientNet fusion; diagnosed a branch-dominance failure suppressing the decoder’s contribution to AUROC and proposed an auxiliary-loss correction.
  • Stacked the base models with scikit-learn and XGBoost meta-learners trained on exported embeddings; resolved solver convergence failures through pipeline-level feature scaling.
  • Integrating RAD-DINO, a chest X-ray–pretrained ViT, as a tenth base model and replacing per-class weighted BCE with Asymmetric Loss.

Summer 2024

Summer Research | Hofstra University Summer Science Research Program

Project: Worked with a student team on a project to repurpose scrap polycarbonate plastic through 3D printing. (Mentor: Dr. Edward Segal)

Research Interests

  • Medical AI & Deep Learning
  • Computer Vision for Imaging
  • Robotics & Autonomous Systems
  • Education Technology

Robotics

My Journey

My journey in FIRST began in FLL and evolved into FTC through the Forest Hills Robotics League. I have stayed in the FIRST community for years, growing from student programmer to team captain.

My Team

From 2023 to 2026, I was captain and lead programmer of team #8721 Mighty Mechanics, a community-based FTC team that represents NYC in high-level competition and outreach.

My Skillset

As lead programmer, I led development of Mighty Motion, our custom Java motion-control library, along with the localization and AutoAim systems built on it. The portfolio covers how it works.

Awards

  • FTC NYC Champion — 2026
  • FTC Control Award — 2026
  • FIRST Tech Challenge World Championship Qualifier — Houston, 2024 and 2026
  • FTC NYC Finalist — 2025
  • FIRST Dean’s List Semi-Finalist, NYC — 2025
FTC World Championships

The Mighty Mechanics and our robot at the 2024 FTC World Championships.

Working on robot arm

Working on our robot's robotic arm.

Our 2026 NYC FTC Championship trophies, medal and Control Award

Our 2026 NYC Championship trophies and Control Award.

Leadership & Activities

Mighty Mechanics FTC Team

Captain & Lead Programmer for FTC Team #8721 Mighty Mechanics (2023 - 2026).

NYC Champion and FTC Control Award winner (2026); World Championship qualifier (2024, 2026).

Stuyvesant ARISTA

Vice President of Web, elected (2025 - 2026); Web Committee Member (2024).

Led a 15-person web team maintaining stuyarista.org and its backend for 200+ members and 1,000+ tutees.

PS 117Q STEM Outreach Program

Program Lead & Instructor (2022 - 2026).

Led a four-year STEM outreach program at a NYC public elementary school, building the lesson plans and running hands-on coding and robotics sessions for 250 students.

Hands-on robotics workshop for 4th grade students

Gocogni.io

Co-Founder & CTO (September 2024 - October 2025).

Demo of the Cogni study platform.

Volunteer Experience

Stuyvesant ARISTA - VP of Web

Led a 15-member web team maintaining stuyarista.org for 200+ members and coordinated policy changes through executive council work.

PS 117Q STEM Outreach Program - Program Lead & Instructor

Ran coding, math, and robotics sessions for 2nd, 3rd, and 4th graders from 2022 to 2026, building the lesson plans for the program.

CS Peer Tutor - Stuyvesant High School

Tutored fellow students in computer science.

National Museum of Mathematics - Volunteer

Volunteered with MoMath in New York City.

Teaching robotics

Teaching 3rd graders robotics concepts with our FTC competition robot.

Teaching coding

Teaching 4th graders coding and algorithm fundamentals.

Teaching code

Teaching basic Java.

ARISTA

Induction ceremony as an ARISTA VP.

Education & Academic Programs

Expected Graduation: May 2030

Carnegie Mellon University

B.S. Information Systems, Computer Science (additional major)

Coursework: Principles of Imperative Computation (15-122), Concepts of Mathematics (21-127), Reasoning with Data (36-200)

Graduated: June 2026

Stuyvesant High School

New York, NY | GPA: 4.0 | National Merit Commended Student

CS electives: Software Development, Systems Programming

Programming contests: PClassic, University of Pennsylvania (2024, 2025); Cornell University High School Programming Contest (2023, 2024, 2025)

11th and 12th Grade

Columbia University Science Honors Program (SHP)

Coursework: Introduction to Algorithms, Introduction to Quantitative Biotechnology, Classical and Quantum Computing, Python Data Structures

Summer 2024

Columbia University SHAPE

Competitive Programming & Machine Learning

Presenting my final project for SHAPE

Presenting my final project for SHAPE

Technical Skills

Grouped by area, with the projects where I've used each.

Machine Learning

Chest X-ray ensemble research

PythonPyTorchTensorFlowDatabricksscikit-learnXGBoostNumPypandas

Robotics

Mighty Motion, ProfiledPathGenerator

JavaFTC SDKKalman filtersPure pursuitPIDSplines

Web Frontend

stuyarista.org, Cogni, NerdyMap, IKEAlyzer

TypeScriptJavaScriptSvelteKitReactTailwindD3.jsBootstrap

Backend & Data

stuyarista.org, Cogni, NerdyMap, IKEAlyzer

FlaskHono.jsNode.jsPocketBaseSupabaseSQLiteSQLCloudflare Workers

Systems & Graphics

c_chat, a Unix shell, the 3D graphics engine

CC++POSIXSocketsGTK4JavaLinuxMake

Mobile & Tools

GoTravelAI, team workflows

SwiftSwiftUIGitGitHubBash