The problem

Traditional solar system models are fixed — they show one configuration and can't be adapted to focus on a single planet or its satellite. For STEAM education, this limits how teachers can use them interactively in class.

The goal was to design a system that is modular: the structure physically adapts depending on which planet is being studied, and each planet's orbital mechanics can be explored in isolation.

Mechanical design

Each planet module uses a ring gear driven by a servo motor mounted inside a compact housing. The satellite arm rides on the ring and can be swapped or repositioned based on the planet's known orbital data. Sub-assemblies were designed to be independent — a planet can be removed and replaced without disassembling the whole system.

All structural components were modeled in Autodesk Inventor, then fabricated using a combination of 3D printing (PLA) and laser-cut acrylic. The ring gear teeth were optimized for the torque range of the servo motors used.

Electronics & control

An Arduino microcontroller manages 4 motors simultaneously, each responsible for a different axis of motion: the planet's orbital rotation, the satellite's orbit, the planet's axial tilt, and the ring's revolution speed. Speed ratios between motors were calculated from real astronomical data to produce accurate relative motion.

Final prototype

The assembled system demonstrates orbital mechanics at classroom scale. The blue sphere (Earth) orbits within the ring assembly while the satellite (Moon) orbits around it — all driven by the motor stack below and housed in a 3D-printed enclosure.

My contributions

• Full mechanical design in Autodesk Inventor — ring gear, housings, modular brackets
• Fabrication: 3D printing all structural and gear components in PLA
• Fabrication: laser cutting the base and mounting plates in acrylic
• Arduino firmware — 4-motor synchronization with astronomical speed ratios
• Written thesis documenting design decisions, trade-offs, and educational applications

Context — the Seaguard Project

The Seaguard Project at USN focuses on developing autonomous systems for maritime monitoring and security. A key challenge is deploying mixed fleets of vehicles — some on the surface, some underwater — that need to work together without constant human oversight.

This thesis sits within that broader effort, focusing specifically on the coordination layer: how do heterogeneous agents with different capabilities, speeds, and sensor ranges divide and cover an area efficiently?

Research focus

The work analyzes existing coverage path planning algorithms and evaluates their suitability for heterogeneous fleets — where agents have fundamentally different motion constraints and sensing capabilities. Key questions include: how should task allocation change when one agent can cover 3D space (underwater) and another is surface-constrained? How do communication limitations affect coordination robustness?

My contributions

• Literature review of coordination strategies for multi-agent and swarm robotics systems
• Analysis of coverage algorithms adapted for heterogeneous maritime platforms
• Simulation and evaluation of coordination approaches in marine environments
• Research supported by the Seaguard group at USN with direct supervision

This thesis is ongoing — this page will be updated as results are finalized.

The problem

Parkinson's disease is typically diagnosed years after neurodegeneration has already begun. Two of the earliest detectable symptoms are changes in grip strength and finger-tapping speed — both measurable non-invasively with the right hardware.

Vitacheck aims to make these clinical tests accessible outside a hospital setting, enabling earlier screening and longitudinal tracking by patients and clinicians.

How it works

The system has two main components: a 3D-printed grip force sensor housing (using an MA-100 force cell) that the patient squeezes, and a tablet running the Vitacheck app. A second test asks the patient to tap a screen as fast as possible over a fixed interval — a digitized version of the standard clinical finger-tapping test.

Both signals are processed in real time, with metrics extracted (peak force, symmetry, tapping rhythm, inter-tap interval variance) and displayed live in the GUI for clinician review.

My contributions

• Integration of MEMS-based force and motion sensors with embedded hardware
• Full Python GUI development for real-time patient interaction and test administration
• Signal acquisition pipeline — from sensor output to processed metrics
• Cross-layer integration: hardware, firmware, and application software
• System design with clinical workflow as the primary constraint

The problem

AKAD Seguros had over 10 years of operational data spread across 6 disconnected systems. In 2022, leadership set a new strategic direction: build a modern data platform to enable data-driven decision-making across all business units — and do it without disrupting live production systems.

Architecture

The platform follows a three-zone data lake pattern: Raw Zone (exact source replicas, append-only), Staging Zone (cleaned, deduplicated, type-corrected), and Curated Zone (business-ready tables ready for analytics and actuarial models).

Data ingestion from relational databases was handled by AWS DMS using Change Data Capture (CDC), ensuring the data lake stays in sync with source systems in near real-time. All processing runs on Amazon EMR using Apache Spark with Delta Lake — enabling ACID transactions, upserts, schema evolution, and time-travel queries across the full dataset.

Outcomes

The platform gave every business unit at AKAD access to clean, versioned, queryable data for the first time. Power BI dashboards and actuarial models were built directly on the Curated Zone. One measurable outcome: a reduction in cargo insurance loss ratios, enabled by better data visibility into claims patterns.

The project was featured in an official AWS Brazil blog post, where I am credited as co-author alongside the AI Engineering Manager and AWS Solutions Architects.

My contributions

• Full architecture design — three-layer lake schema, zone boundaries, and data contracts
• AWS DMS setup for CDC ingestion from 6 relational database sources
• Delta Lake implementation on Amazon EMR with ACID, upserts, and time-travel
• Pipeline orchestration with AWS Step Functions
• Data Warehouse build on top of the Curated Zone for Power BI and actuarial teams
• Monitoring and observability with AWS CloudWatch across all pipeline stages
• Data catalog with AWS Glue across all three layers

The challenge

The Trinity College Firefighting Robot Contest places autonomous robots in a house-like maze arena. When an alarm triggers, the robot must navigate from its starting position, locate a lit candle (representing a fire), and extinguish it — all without human intervention.

The arena is unknown ahead of time, so the robot must map and navigate dynamically, detect the flame's direction and proximity, and deploy water precisely without damaging electronics.

Design & engineering

I led the structural design and CAD modeling in Autodesk Inventor, translating competition constraints into a compact chassis that could house sensors, a water pump, and drive electronics. Key components were fabricated using 3D printing and aluminium profile framing.

The embedded logic ran on Arduino: sensor fusion from multiple ultrasonic sensors for wall-following navigation, flame sensor array for detection and direction, and motor control for drive and the extinguisher mechanism.

My contributions

• Full robot structure modeled in Autodesk Inventor (CAD)
• 3D printing of custom structural and mounting components
• Arduino-based embedded logic: sensor input, navigation, flame detection, and motor control
• Water containment solution design — sealed bird feeder repurposed as reservoir
• Integration and system-level debugging ahead of competition