Rayhan Rinzan

BS in Computer Science @ Cornell University

Rayhan Rinzan

BS in Computer Science @ Cornell University, 2025 - 2029. Ithaca, NY.

Exploring opportunities in software engineering, machine learning, and data science.

Where I've been

  1. World Resources Institute (via Develop for Good)

    • Software Engineer to

    I’m working on Energy Access Explorer, an open-source geospatial platform used by governments and organizations to plan electrification and clean energy access. I’m helping build features that improve how users explore and evaluate energy data, with contributions intended for the platform’s public codebase.

  2. IMGNi

    • Software Engineer Intern to now

    I’m building Sienna AI, an AI employee for small businesses.

  3. Cornell University AutoBoat Project Team

    • Software Engineer to now

      I write perception software for Cornell's autonomous boat. I added quaternion-derived yaw to the exponential moving averages used for vessel tracking, and buoy color and sign validation, so the autonomy stack gets more consistent information about tracked objects.

    • Electrical Systems to now

      Contributed to Robotics Power Distribution + Binary Relay Board

  4. Weill Cornell Medicine

    • ML Research Intern to

    I built and evaluated deep learning pipelines that segment abdominal organs in MRI scans for radiotherapy planning. I worked on the dataset infrastructure, preprocessing and GPU training in Python, PyTorch and Linux, across hundreds of experiments. The segmentation system reached about 0.94 Dice.

  5. Syracuse University College of Engineering and Computer Science

    • Robotics Systems Engineer to

    I worked on a radar-equipped hexacopter built on Pixhawk, ArduPilot and a Raspberry Pi, on sensor integration and synthetic aperture radar processing. I wrote image-reconstruction workflows that use motion compensation and backprojection to account for the drone moving while the radar records.

  6. SUNY Oswego

    • Computational Research Intern to

    I wrote Python tools to analyze experimental data for research on ladder-type electromagnetically induced transparency in cesium. They were used to study the system’s optical response and to help interpret the measurements.

Things I've made

Keepsake, a wearable memory aid for people with Alzheimer’s and dementia

Built at HackHarvard 2026 with a four-person team for people living with Alzheimer’s and dementia. A chest-worn iPhone streams to a laptop running a real-time vision pipeline at 5 fps: an open-vocabulary YOLO-World detector alongside two YOLO models we fine-tuned on our own items, a CLIP second opinion that rejects each item’s known look-alikes, and MediaPipe hand tracking as evidence of pick-ups and put-downs. A sighting only counts after three detections within a second, so one false frame never becomes a memory. Ask “where are my keys?” and it answers with when it last saw them, a Grok vision description of what was around them, and the photo. My focus was the reminder system, built to help without nagging: when an item you had stays out of view, it sends one gentle iMessage through Photon Spectrum with the item circled in the photo, worded as “last seen”, never “you forgot”. Per-item and global cooldowns cap how often it can text, replies like “I have it” or “remind me later” settle it, and a reminder is dropped if the camera goes dark or freezes, then checked again just before it’s sent. Support scales with the person: High-Assistance mode texts about ten seconds after any steady sighting goes out of view, Everyday mode waits minutes and acts only on a confirmed put-down, and the app’s three help levels range from listening only when tapped to checking in on its own.

Python, FastAPI, YOLO, CLIP, MediaPipe, Expo, React, TypeScript, Photon Spectrum, Grok API, SQLite, Vercel

FirstTouch, arcade 4v4 football in first person

A first-person 4v4 football game that runs in the browser: bend curlers around the keeper in slow motion, double jump into headers, catch the ball mid-air in bullet time and fire knuckle-ball super shots. The hard part was the ball physics, written from scratch with no physics engine: ballistic flight with air drag, spin-based curve and knuckle-ball wobble, and bounces off walls, posts and nets, with fast shots substepped up to 8 times a frame so they can’t tunnel through posts or players. Passes are lofted with a closed-form solve for gravity and drag, so they land where they’re aimed, and teammates iterate on where you’ll be when the ball arrives to lead passes onto your run. Each bot runs a state machine (carry, shoot, pass, cross, chip, chase, support), and the keepers position, dive, rush out and can be fooled by curl. It is a single HTML file with no image or model assets; the pitch, players and crowd are generated in code, and every sound is synthesized live with Web Audio.

Three.js, JavaScript, WebGL, Web Audio API, game physics, game AI

QuoteDesk, a live manufacturing quoting assistant

Built for Google DeepMind New Agentic Economy 2026 with a three-person team. It turns a buyer’s email and engineering drawing into a provisional quote by asking for missing details through Gemini Live. My focus was connecting the voice conversation to the quoting system. The hard part was making spoken answers reliable state changes: schema-defined tool calls pass through server-side validation, version checks prevent conflicting updates, and duplicate-call protection keeps retries from applying an answer twice. Supabase stores the shared specification and event history, while Realtime subscriptions and a polling fallback update the estimator dashboard. Buyers can speak or type, correct earlier answers to trigger deterministic repricing, and request discounts that are flagged for vendor review rather than automatically accepted.

Next.js, TypeScript, React, Gemini Live API, Supabase, PostgreSQL, Zod, Vercel

FIFA World Cup ML forecasting engine

It turns historical international football results into match predictions and tournament forecasts. The hard part was draws. My first models reached about 53% accuracy by never predicting one: zero draws called across 8,155 test matches. So I stopped scoring on accuracy, trained with class-balanced weights, and judged every model by macro F1 on a chronological split, where ignoring an outcome costs a third of the score. Adding leakage-safe Elo ratings then lifted macro F1 about 18%, from a 0.445 baseline to 0.526, with a model that predicts all three outcomes. Knockout games can’t end level, so the bracket simulator splits each draw probability between the two teams before a Streamlit app runs 10,000 Monte Carlo tournaments to estimate every team’s chance of reaching each round.

Python, scikit-learn, pandas, Jupyter, Streamlit

ThirdMan, an AI football tactics sandbox

Ask it a tactics question in plain language and it animates the answer on an editable 22-player pitch. The hard part was the simulation engine behind the model’s structured output: it models how defenders react, checks passing lanes and interceptions, and searches for supporting runs. Playback is deterministic, with pause, scrubbing and replay, plus opponent-response comparisons and undo/redo. It still works in full when the AI service is unavailable.

Next.js, TypeScript, React, SVG, Motion, Zod, OpenAI API, Vercel

ClockedIn, a calendar-aware productivity app

An iOS app that shows you how you spend your time, through periodic check-ins, an activity log and daily productivity charts. The hard part was the reminder scheduling: it reads the calendar and skips quiet hours and the events you pick. Notification actions open an activity entry that is already filled in.

Swift, SwiftUI, EventKit, iOS notifications

skillissue.lol, a 100% accurate internship recruiting simulator

An internship recruiting simulator written in LOLCODE, because I was a little bored. It screens your experience and GPA, runs a technical interview where the right answer to “production is down” is to blame the intern, and then rejects you no matter what. A perfect score gets flagged as AI-generated; anything less misses the required 101% match.

LOLCODE, HTML, JavaScript, GitHub Pages

Where I study

Cornell University

to

BS in Computer Science

Coursework
Data Structures and OOP, Functional Programming, Discrete Math, Data Science and Decision Making, Linear Algebra, Differential Equations, Circuits
Activities
Autonomous Boat Project Team (Software), Institute of Electrical and Electronics Engineers (PR Team), South Asian Council (Publicity Team), Society of Asian Scientists and Engineers, AI Alignment, Muslim Educational & Cultural Association, Notion Campus Leader, Adobe Campus Ambassador

What I've won

  • 1st place, MIT BeaverWorks 2025

    UAS-SAR Signal Processing (International Division)

  • 2nd place, Lockheed Martin CodeQuest

    Advanced Division

About me

I'm Rayhan, a computer science student at Cornell University interested in intelligent systems, with a focus on machine learning, autonomy, and data engineering. My experience spans software engineering, robotics, and applied machine learning.

Say hi