Selected Engineering Systems

Projects

Systems I've built across healthcare AI, embedded firmware, robotics, hardware validation, and full-stack platforms — with emphasis on real constraints, measurable outcomes, and production judgment.

View Experience Contact Me

Featured Work

Systems with real-world constraints

A focused selection of projects showing how I move from problem definition to implementation, validation, and impact.

01 / Healthcare AI

CareFuse

Co-founder and ML Lead

Built a clinical AI platform for predicting orthopedic treatment outcomes, comparing surgery against conservative care, and generating explanations for payer/provider workflows.

Designed ML-ready datasets, calibrated models, SHAP explanations, and validation workflow.
Built FastAPI + Docker deployment path for model serving and internal tooling.
Supported pilot discovery with a Brazilian healthcare organization serving 10M+ covered lives.

Python scikit-learn XGBoost SHAP FastAPI Docker

Read CareFuse case study

Evidence

AUC ≈ 0.93, calibrated: Validated outcome models with calibration and SHAP explanations.
FastAPI + Docker: Reproducible deployment path for model serving and internal tooling.
10M+ covered lives: Pilot discovery with a major Brazilian healthcare organization.

02 / Embedded & Hardware Validation

Automated Motor Controller Validation

Electrical Engineering Intern, DeWALT

Designed and built an embedded test fixture for DeWALT motor-controller validation, replacing slow manual testing with a repeatable automated workflow.

Designed the PCB in Altium and wrote MSP432 firmware with I2C drivers and thermocouple sensing.
Built fixture electronics, enclosure wiring, logging, and a simple operator workflow.
Cut test time by roughly 4 hours and improved repeatability for hardware validation.

MSP432 Altium I2C Thermocouples C firmware

Custom validation PCB designed in Altium for DeWALT motor-controller testing — Custom validation PCB designed in Altium for the automated test fixture.

03 / Robotics & Electrical Systems

Autonomous Hydrofoil RHIB Platform

Electrical System Lead, CMAR

Designed the electrical backbone for an autonomous hydrofoil RHIB, connecting power, control, sensing, and autonomy into a maintainable system.

Specified motors, batteries, throttles, microcontrollers, and power components.
Built a CAN backbone with sonar parsing and ROS drivers across subsystems.
Moved the RHIB toward a robust autonomous platform ready for real-world testing.

CAN ROS Sonar Power systems Marine hardware

RHIB electrical architecture diagram showing port, starboard, brain, batteries, CAN, USB, safety, and actuator subsystems — Electrical architecture: port, starboard, brain, batteries, CAN, safety, and actuator subsystems.

04 / Full-Stack Platform

Spurr

Founder and Lead Engineer

Built a campus platform that helps students find organizations, events, and verified group information without scattered chats or stale calendars.

Designed auth, role-based access control, claims, verification, and moderation flows.
Built searchable organization and event data on a Next.js + Supabase stack.
Modeled 160+ organizations, groups, and events in a deployed product.

Next.js TypeScript Supabase PostgreSQL Vercel

Visit Spurr

Evidence

160+ groups & events: Organizations, groups, and events modeled in a deployed product.
RBAC & moderation: Claims, verification, and role-based access control built in from the start.
Shipped on Vercel: Next.js + Supabase stack running live at spurr.life.

05 / Humanoid Robotics

TREC Humanoid Robot Pandora

Research Fellow, Terrestrial Robotics Engineering & Controls

Hardware, firmware, and sensing integration for Pandora, a humanoid robotics research platform at Virginia Tech, so higher-level control work could be tested on reliable electronics.

Designed digital circuits and a sensor shield PCB to interface sensors with the embedded stack.
Updated firmware connecting embedded hardware, data acquisition, and robot-level functionality.
Worked across board-level design, firmware, and robotics constraints to improve sensing reliability.

C/C++ Embedded Firmware PCB Design Sensor Integration Data Acquisition Robotics Hardware

Scope

Sensor shield PCB: Board-level integration linking sensors to the robot's embedded stack.
Firmware & DAQ: Firmware updates moving sensor data into the robot's control software.
Hardware/software co-design: Decisions made across board, firmware, and robotics layers for reliability.

TREC Humanoid Robot Pandora research poster — Research poster for Pandora. Click to view full size.

Additional Builds

Other Technical Projects

Smaller builds across probabilistic robotics, embedded sensing, signal processing, and edge compute.

Probabilistic Robotics

Bayesian Obstacle Detection for Underwater Robots

A Bayesian obstacle detection and avoidance system for underwater robots that fuses noisy sonar returns into a probabilistic map. Each measurement updates belief about nearby obstacles instead of forcing a single yes/no decision, making navigation more robust to uncertain sensing.

Bayesian Inference Sonar Probabilistic Mapping Obstacle Avoidance Python ROS

Edge AI

Edge AI Low-Power Drone Detection

A low-power sensing pipeline for real-time drone and noise detection on a Raspberry Pi Zero 2W.

Python Edge AI Signal processing Embedded Linux

Analog Circuits

Infrared Radioteletype

An IR communication system using optical transmission, analog filtering, and embedded control.

Arduino Photodiode Butterworth filters Analog circuits

Embedded Linux

Smart Home System

A PCB-centered edge compute concept tying Raspberry Pi GPIO, power breakout, storage, and automation hardware together.

PCB Raspberry Pi GPIO NVMe Edge compute

Approach

How I Build

The same pattern runs through these projects, whether the deliverable is a model, a board, or a platform.

Define the system constraint

Start from the real limit — accuracy, current, latency, or trust — that decides whether the system is usable.

Build the minimum reliable path

Ship the smallest end-to-end version that actually works before adding scope or polish.

Validate with measurable outcomes

Prove it with calibration, test logs, or repeatable benchmarks — not anecdotes.

Package the work so others can use it

Wrap it in APIs, fixtures, docs, or tooling so the result outlives a single demo.

Next: Academics