Designed and synthesized a pipelined LC2K processor in SystemVerilog with an out-of-order micro-op queue, banked BTB, and direct-mapped Icache, achieving 1.6 CPI and 18.9 ns per instruction across the full test suite for a top-5 class ranking.

Built bare-metal C firmware on an STM32 ARM Cortex-M4 (no RTOS) integrating 6+ peripherals over SPI, I2C, and UART for wireless gesture-controlled audio recording and playback, with an interrupt-driven gesture recognition pipeline mapping sensor input to audio events.

Developed and shipped a 2D platformer in Unity (C#) themed around 3D printing concepts for the S2A Research Lab, implementing physics-based mechanics like coyote time and responsive jump; demoed to the lab team and at the SURE Symposium.

Implemented a feedforward neural network from scratch in C# — custom layer/matrix math, sigmoid activation, and forward propagation with no ML libraries — then paired it with a genetic algorithm that mutates weights and biases each generation to train simulated swarm robots in Unity.

A Next.js/TypeScript app backed by Prisma and Supabase that tracks everyday purchases (food, soap, toothpaste, replacements) to compute survival spend per day, week, month, and year; deployed in invite-only alpha.