Coding

Fjord is a business-automation platform for home-service contractors: white-label lead-capture websites, automated Google review requests, instant 24/7 follow-up on missed calls and texts, and a lead dashboard tracking every interaction - built so a contractor can go live in about a week.

I architected the multi-tenant SaaS backend behind it - a Node/Express and Supabase API with 59 REST endpoints across 13 database tables, isolating every tenant's dashboards, pipeline, and site - and built the Gemini-powered AI layer that qualifies inbound SMS leads and drafts personalized replies in each business's voice.

Axia Prep is an AP exam-practice platform for full-length tests, unit quizzes, FRQ and MCQ drills, and performance analytics.

I built the course-material ingestion and generation pipeline behind its structured AP-style question banks, including diagram formats and grading that retains the relevant unit and exam context.

M.Eng. HCI prototype for digital literacy on social posts: upload an image, optionally add a caption or claim, and get plain-language scores for misleadingness, misalignment, and how likely the image is AI-generated.

Multimodal checks pair vision with GPT-4 text reasoning so non-experts can see what the model noticed—not just a raw confidence number.

Daily Market Analysis is a Python Shiny dashboard that pulls live stocks, forex, commodities, news sentiment, and economic indicators from the Alpha Vantage API into one interactive workstation - with scenario projections, position calculators, and multi-chart analytics per section.

Beyond the dashboard, it runs a three-agent AI pipeline: an Orchestrator that plans the analysis, a Market Analyst that writes an evidence-based memo, and a Lead Editor that condenses it into a clean brief. Retrieval-augmented generation grounds every report in a custom market-notes corpus, keeping the AI tethered to real data instead of guessing. Built for resilience - cached fallbacks when API limits hit, automatic LLM provider failover - and deployed on Render.

Scheduled message delivery in Zulip.

INFO 4940 studio project: Be My Cupid / DAM, an AI matchmaker that turns your preferences into three fictional dating profiles—a close match, a moderate stretch, and an exploratory pick—then uses swipe feedback to refine the next round.

Built around intentional disclosure and shared data rather than endless swipe chrome. Preferences stay in the browser; Gemini generates fresh profiles each session with no real-person database.

RescueSight is a prototype AI-powered emergency response system for smart glasses, built over a weekend at the Cornell AI Health Hackathon. A voice agent asks simple questions to assess the situation while a computer vision model reads visual cues from the scene - like hand placement during CPR - so the system can identify what's happening and guide a bystander through it step by step.

AI presentation APIs and brand-safe deck agents.

Rental listings with review credibility scores.

Fault-tolerant sharded key-value store.

ROS localization, planning, and control.

MyResLife is the residential-life app for Alice Cook House at Cornell—house events, dining check-in, mail, and day-to-day house communications in one place. I built the Flutter client with CampusGroups auth and live Firestore-backed hall data, and we shipped it to the App Store for students, staff, and faculty in the house.

Improving ECE 2300 labs, CI, and a custom student linter.

EfficientNet and a bottom-up deep learning path.

Generative models studio.

AI class schedule builder.

Network topology meets training parallelization.

Contributed to CAST, Cornell CALS' Farm of the Future testbed for data-driven dairy and crop research - piping live IoT sensor readings into Postgres and building a React dashboard that flags gaps and spikes before they become bad data, plus simple models to catch flaky sensors.

Startup hangout recommendations for groups in Jordan.

Genre runtimes before and after 2010.

BERT across 28 sentiment categories.

Fellowship with a human-centered startup.

Volumetric data, visualized.

Timed Wordle in the terminal, in OCaml.

A full processor path in RTL.

Summer counselor at Summer at King's.

I built The Cornell Dining Forum in Spring 2023 as part of Cornell’s INFO 1998 Trends in Web Development course. The app lets students rate dishes at campus dining halls and browse what other students are saying.

I used the project to get hands-on with React and Next.js for the frontend, and Firebase Auth and Firestore for sign-in and storing reviews. Working through that stack helped me connect classroom concepts to a real, deployable web app.

GreenBin (Big Red Hacks 2022, Cornell) is a campus habit tracker for recycling, zero-waste, biking, and walking. The idea was to verify actions with photos: Google Cloud Vision would read labels from the upload and suggest an activity, then feed timed challenges and a community board so progress was shared, not private.

Built with Mena, Nada, and Youssef Attia.

Yal3ab (يلعب), the Arabic word for "play," features two games: my custom-designed Funny Animals Memory game and my friend's Hangman game, both built while we were learning React. This was one of my first coding projects, and we wanted to create something anyone could open, play, and compete on through a shared scoreboard.

During the COVID-19 pandemic, thousands contributed personal journal entries to the Pandemic Journaling Project a project organized by the University of Connecticut's Department of English to create a repository of our ownunique records of everyday life during the pandemic.

I wanted to explore whether computational text analysis tools could help us discover patterns that are impossible to notice by reading entries one at a time. Using Mathematica and Wolfram Language, I built an analysis notebook and explored things like word frequencies, contextual bigrams, sentence relationships, and temporal distributions to study how emotions, routines, and perceptions evolved over that time.

I focused on the context surrounding words to better understand the collective human experience. The project showed what we called in that class 'distant reading' which can transform thousands of individual stories into insights that we would never be able to come at by reading entries one at a time.

I spent a high-school internship at TechWorks (now The Makerspace), a Crown Prince Foundation digital fabrication lab in Amman built for young innovators to prototype with advanced tools - 3D printers, laser cutters, CNC mills, and more. I wasn't shipping a product of my own. Day to day I helped college students with their designs: CAD and CAM walkthroughs, choosing the right machine, and coaching them on the shop floor until a sketch became a part.