Making and Side Projects

My first big project in the PROOF Lab: a base for the Doosan H2515 to be mounted on. It is built from 8020 extrusions, sized against the robot's maximum payload so it cannot tip when the arm works out over an edge. I built the base and supervised the installation of the robot onto it.

I started from an STL model of the robot, which drove the layout. Custom plates were needed for the angular truss that resists twist and keeps the frame square; I machined those plates myself in the machine shop and drilled the top plates to accept screws into the 8020s. Over 200 screws went into the assembly. The power supply is fitted underneath, and 45 lb gym plates sit on the base as ballast.

• SolidWorks
• FBD Calculations
• Machining

I developed labs for E311, a materials processing course at Stevens taught by Prof. Sarah Goodman. I synthesised gold nanoparticles by the Turkevich method and tuned their size through the concentration of the sodium citrate solution. Because of their refractive properties, the differently sized particles change the colour of the solution, which makes the result immediately visible to students. The lab runs at under one dollar per student. I presented the findings at the ASEE MAS Fall 2023 Conference.

Gold (III) chloride is diluted with distilled water. Stirred constantly with dilute sodium citrate under heat, the solution turns from yellow to grey and then to red or purple, which marks the formation of the nanoparticles. I varied the heat and mixing duration looking for observable differences in colour or size; the best lever turned out to be sodium citrate concentration, which yielded a spread of about 50 nm.

Sizes were confirmed on a scanning electron microscope I was trained to use — samples pipetted onto an SEM stand with a silicon wafer, then dried properly before imaging. The solution colour was quantified by spectrophotometry, which outputs a wavelength. Together those measurements were enough to establish the relationship between concentration and particle size.

• Scanning Electron Microscopy
• Spectrophotometry
• Molarity Calculations

I needed a study lamp for my room. It is made from basswood and walnut, with the light diffused through shoji paper lined onto the kumiko panels. The lamp is designed so that a broken panel can be detached, serviced, and replaced rather than scrapping the whole thing.

I started by watching videos for inspiration, then sketching out ideas for the panels. This was a major quarantine project. Power tools felt both intimidating and wasteful here — the kerf of a table saw blade rivals the thickness of a single kumiko strip — so I hand cut and planed every piece to size and glued up all the joints.

Spray adhesive was applied to the back to accept the shoji paper, which was carefully laid onto the kumiko and left to dry. Dabs of hot glue hold the panels in the walnut frame, which is what makes them easy to replace.

• Hand-Tool Woodworking
• Patience

A watch box in walnut and poplar. The corners are dovetailed with a miter at the bottom to disguise the slot routed for the base panel. It holds six watches comfortably, and poplar was chosen for the inner liner because it is softer against the cases.

The hardest part was cutting the mitered dovetails at each corner, but they came out pretty nice.

• Hand-Tool Woodworking

I wanted a block plane in brass and steel for my own woodworking. The machine shop at Stevens taught me to run the manual CNC machines and to write and run programs on them, so I wrote the programs that carve out the rounded dovetails and machined the components myself.

The project stays a work in progress until I have the courage to take a hammer to the dovetails and make them permanent. All the components are still separate: about a month of hard work went into machining the body, so I am hesitant to commit the joints in case something goes wrong. I also still need a blade — my custom width is narrower than off-the-shelf irons, so one will have to be made and heat treated to hold an edge.

• CAD
• CNC Machining
• Manual Machining

Virtually Generated Engineered Organic Realistic George E-assistant (haha). I followed a YouTube tutorial on building a home assistant with AI and modified the code for what I wanted from it. A neural network with natural language processing takes vocal commands and maps them to functions. A YAML file holds all the inputs and maps them to outputs, which can themselves be functions that break the input down further — a weather API returning a default temperature you set, or the forecast for a location you name.

The modules I wrote include the weather API, a random number generator, a timer, and a Google search feature. I used Google text-to-speech to make the voice sound less robotic. If I redid this project I would push more generalisability deeper into the layers and add custom apps like notetaking, calendar entries, and event notifications.

• PyTorch
• NLP
• API Calls
• Python

Cutting operations on the floor were being done manually by machine operators, which made for a dangerous work environment. I designed a fixture in Siemens NX and had it built from steel, employing GD&T principles for the drawings; the machinists were enormously helpful in guiding me through machining tolerances.

The final fixture was deployed on the floor, securing the tibial casting safely and making the operation hands-free. It is swappable, so fixtures for other part families can be dropped into the same setup.

• Siemens NX
• GD&T