GWatch Custom Smartwatch
I built a smartwatch from scratch. Hardware, software, and enclosure. It connects to iPhone and shows real-time notifications.
Why I Failed the First Time
I tried this nine months ago and quit immediately. I spent weeks comparing hardware and planning features before building anything. I was optimizing a product that didn't exist. Nothing got built. The lesson: without a physical prototype, every decision is a guess.
What Changed
I forced one constraint: build the simplest thing that works. Three requirements. Nothing else. Connect to iPhone. Display time. Show notifications. Everything else was cut.
Iteration 1: First Working Prototype
iPhone to watch communication: working. Notifications displaying in real time. Clean minimal UI. Built in a single day. The enclosure was 3D printed, press-fit, no screws. No styling. No overengineering. Just something wearable enough to learn from.
Iteration 2: Smaller, Tighter, Simpler
The first enclosure was too bulky. I measured it, identified what was wasting space, and cut it. Converted from two-part to unibody. Fewer parts, no seam, no tolerance stack between components. Added bottom ribs for tool-free opening.
- Height: 20 mm → 16.4 mm (18% reduction)
- Diameter: 46.5 mm → 44.5 mm (4.3% reduction)
Iteration 3: Engineered to the Boundary
Acceptable size wasn't good enough. I rebuilt the enclosure around the components, not around assumptions. Removed the MCU socket entirely. Dropped the microcontroller as low as the battery connection allows. Projected every component outward to the circular boundary so nothing wastes space inside or creates external protrusions. Tolerance set at 0.2 mm throughout. Sharp internal angles replaced with large radii to eliminate FDM print inconsistencies at direction changes. String band replaced with a magnet bracelet. Final geometry will be machined in metal.
What This Shows
Every change is measured. Every decision has a reason. The watch went from idea to functional wearable in a day because scope was controlled, not because the problem was easy. Hardware tested. Software tested. Enclosure being finalized before metal production.
ATLAS Automated Solar Tracking Platform
ATLAS is a roof-mounted single-axis solar tracking platform built for residential homeowners. It combines automated sun tracking, self-cleaning, and real-time monitoring in a system designed to generate 20% more energy than fixed panels and pay itself back in under 8 years. I led the project as team lead across a 6-engineer team, owned the sensor subsystem enclosure from concept to completion, and am currently building a competing elevation mechanism to challenge a fundamental flaw in the current design.
What I Led
Six engineers. One system. My job was to make sure it all worked together. That meant owning the Gantt chart, running weekly team meetings, coordinating across subsystems, and making architectural decisions about how the pieces fit. When the rotation mechanism, controls system, and self-cleaning subsystem all had to function as a single product, I was the one making sure they could. I also handled the unglamorous work: team-building, progress tracking, stakeholder updates, documentation structure. None of it is visible in the final report. All of it determined whether the project worked.
Sensor Module Enclosure
The sensor module is entirely my design. It is the core of the tracking system. The enclosure houses two photoresistors separated by a precision divider wall. The geometry is not arbitrary; it was calculated and then simulated in CAD to define exactly when the system triggers corrective motion.
- Misalignment detection range: 10° to 20°. At 10° the shadow coverage difference becomes measurable. At 20° one sensor is fully shaded.
- Body is PLA. Windows are laser-cut acrylic. Assembly superglued for material compatibility and seal quality.
- Waterproofing test: 50 spray cycles. Zero leakage detected. Four water-sensitive indicators placed inside remained white throughout.
- Latitude robustness confirmed: sidewall geometry maintains reliable shadow discrimination at up to 20° of latitude misalignment, verified through CAD simulation at 75° sun elevation.
Why the Current Elevation Mechanism Has a Problem
The existing rotation mechanism uses two linear actuators mounted side by side. Neither pushes the solar panel at its exact center. This introduces an unintended bending moment that rotates the panel around the wrong axis. The team identified it during first movement testing. It requires a hinge redesign and scissor hinge additions to address. I am not waiting for that fix. I am building a competing mechanism.
Competing Elevation Mechanism: Lead Screw Design
My design eliminates the problem at the source. The lead screw runs inline, centered along the panel's axis. Force is applied at the center, with no offset and no moment. The panel lifts cleanly along the intended axis and nothing else.
- Centering the mechanism provides precise angle control. A stepper motor driving a lead screw gives deterministic positioning. Each step represents a known angular increment.
- The inline layout reduces the part count. One motor. One screw. Significantly more compact than the dual-actuator design.
- If it works as designed, it solves the v1 problem, adds precision the current system doesn't have, and does it with a simpler, smaller package.
Santa Clara University B.S. Mechanical Engineering
Advanced my mechanical engineering work through higher-level design, controls, testing, and simulation courses while building projects that tied coursework directly to hardware and analysis.
Work at Hyster-Yale MECOP Internship
MECOP intern on the Global Modeling & Simulation (GM&S) team at Hyster-Yale Materials Handling, Counter Balance Development Center in Fairview, Oregon. Three major projects: EAGLE (protective enclosure, 98% cost reduction), AWESOME (weld automation, 70-87% time savings), and FEA for C23/P12 adapters (nonlinear structural simulation). Mentored by Max Stein, Engineer IV.
EAGLE Protective Enclosure
Effective Armored Guard Layer Enclosure. Designed, built, and installed a protective enclosure for the lab's Instron 8801 Fatigue Testing Machine. Cost went from $17,646 to $850 (98% savings). Polycarbonate panels deliver 40x the impact strength of the old acrylic design.
I led this project from concept to installation. I interviewed every lab worker about their experience with the old enclosure and collected feedback on what a replacement needed to do. The main design constraint was the machine's physical dimensions. I calculated the optimal height to protect the specimen without interfering with the machine at its lowest position.
- Cost reduction: $17,646 → $850 (98% savings). The Instron-quoted solution cost $17,646 with a 14-16 week lead time. My design provided equivalent or better protection at a fraction of the cost.
- Polycarbonate panels provide 40x impact strength vs the old acrylic design. Researched all transparent panel materials and selected polycarbonate for its superior protective properties.
- Single door with lift-off hinges. The old cover required the user to lift the entire enclosure on and off each time. My solution uses one door for daily access and lift-off hinges for removal when oversized specimens need testing. Red-colored handles distinguish the door operation from the lift-off action, an intentional UX detail.
- 8020 aluminum extrusion frame chosen for low cost, high customization, and simple assembly. Created full BOM, received contractor quote, and installed the final enclosure on the machine.
- Designed in Siemens NX. Protection covers the full range of machine operation with item falling prevention built in.
AWESOME Weld Automation
Automatic Welds Efficiently Saving Operational Minutes Effectively. Investigated and confirmed a new ANSYS Mechanical 2024 Weld function, achieving 70-87% time reduction on weld modeling and welding a 150+ weld lift truck frame in 3 days vs 2 weeks manually.
GM&S had a significant backlog of FEA requests and limited resources. Modeling welds for durability analysis was critical but extremely time-consuming. The old workflow required manually creating weld CAD geometry, welding it to the target body, meshing, troubleshooting, and only then running Seamweld Analysis in Ncode. The new ANSYS Weld tool generates welds automatically, both intermittent and continuous seams, with configurable properties like leg length, thickness, and material.
- 70-87% time reduction on weld modeling. Created a time-savings study with three cases: manual welds vs automatic. Best-case scenario showed 87% time savings.
- 150+ weld frame completed in 3 days vs 2 weeks manually. An average 2-3 ton lift truck frame has over 150 welds. I welded a full frame using the automatic function in 3 days, a 70% decrease in time.
- Authored the Automatic Welding Procedure document for the team, covering theory, procedure, limitations, best usage cases, and troubleshooting of commonly faced issues.
- Enabled GM&S to apply Seamweld analysis to a much bigger variety of projects, and allowed new team members to perform quick draft weld analysis to identify points of interest before manual refinement.
FEA C23 Wheel & Adapter, P12 Adapter
Structural FEA for the C23 Wheel & Adapter and P12 Adapter for the Wheel Force Transducer. Nonlinear simulations with bolts, preloads, and frictional contacts. All tested parts showed stresses below endurance limits.
- C23 Wheel Nonlinear Static Structural Analysis. One of the most accurate simulations I ran. Included all bolts, preloads, assumed base, and frictional contacts. Results showed stresses below endurance. The part passed testing.
- C23 Adapter 0.35 mm radius issue. A critical radius existed on the physical part but not in the CAD model. The instrumental lab measured it, and I modeled a 0.35 mm radius with high-density mesh at points of interest. Linear static structural analysis confirmed maximum principal stress was below the endurance limit of the material.
- P12 Adapter Three design revisions. First design met stress requirements after three revisions but was rejected on cost. Second design (two parts) was fully simulated as nonlinear with all bolts, preloads, and frictional contacts. It did not meet stress requirements. Third design was in progress at internship end.
Mechanical Design Engineer (Student Team) Portland State Aerospace Society
Worked on a student rocket team supporting mechanical design work, learning how to translate concepts into hardware that could be built, tested, and improved for flight-related systems.
Portland State University Mechanical Engineering (Transfer coursework)
Built the foundation of my mechanical engineering track through transfer coursework, lab-based problem solving, and early design projects that shaped my hands-on engineering approach.