Aria Zadeh

Founded a local chapter focused on advancing STEM education for younger students. Led hands-on lessons and mentored students through engineering and science projects, breaking down high-level concepts into accessible and interactive learning experiences.

Science Olympiad - Event leader for Robot Tour and Electric Vehicle - Robot Tour: Programmed a autonomous navigation robot using IMU heading data and motor encoder feedback to correct drift and control distance, writing custom algorithms to achieve high positional accuracy across complex grid courses - Electric Vehicle: Implemented a PID control loop and coded the vehicle to automatically account for encoder values to hit precise distances and timing targets consistently Dawgma FRC 1712 - Control Systems - Wrote robot subsystems and command-based architecture in Java using the WPILib framework, structuring code around hardware abstraction layers for clean separation between subsystem logic and commands - Built and tuned autonomous routines in PathPlanner, configuring trajectories and event markers to execute precise multi-step sequences during the autonomous period - Coded a driver-assist feature using odometry to calculate the robot's field-relative position, identify the nearest scoring target geometrically, and automatically navigate and align for a score - Handled wiring and electrical integration including wire crimping, CAN bus management, and sensor wiring for encoders and other devices

Competed in the MIT Beaver Works CRE[AT]E Challenge, a national program where high school teams engineer assistive technology solutions alongside real co-designers with disabilities. Led development of Focal, a task management webapp built for a high school student with ADHD. Designed and built the full app, deployed to web via Vercel and distributed directly to the co-designer for testing. Includes an AI task breakdown system using Google Gemini, a smart scheduling algorithm that learns from user behavior over time, and a focus session mode targeting initiation paralysis. Tech stack: React Native, TypeScript, Supabase, Google Gemini 2.5 Flash, Vercel.

- Built a modular Python pipeline to ingest and harmonize large-scale NHANES biomedical datasets across multiple survey cycles, handling automated cleaning, merging, and imputation - Designed a flexible configuration system allowing model architectures (XGBoost, Random Forest, Gradient Boosting, etc.) to be swapped without touching core pipeline logic - Implemented and compared multiple class imbalance strategies including SMOTE, ADASYN, Borderline SMOTE, and a custom hybrid undersample-then-oversample pipeline tuned with k-nearest neighbor parameters - Built an experiment orchestration system that programmatically generated isolated config files, executed runs in sandboxed directories, and logged results to CSV and JSONL leaderboards tracking macro F1, recall, and precision across hundreds of configurations - Project focused on binary classification for predicting chronic liver inflammation using elastography measurements, inflammatory biomarkers, and metabolic indicators from NHANES data