James Farmer — Mechanical Engineer & Robotics Researcher

TIKO V4c is the prosthetic-focused evolution of the TIKO platform. Mechanically, its primary innovation is weight and size reduction: smaller motors and tighter packaging shrink the forearm to roughly one-quarter the volume of TIKO V4 while incorporating two additional motors. A new differential wrist adds two degrees of freedom, enabling the combined flexion/extension and pronation/supination motions critical to advanced prosthetic function.

On the software side, V4c introduces the TIKO control system, which addresses a key dichotomy in prosthetics: a high-DOF mechanical hand is only as useful as the control layer driving it. Instead of manual grasp selection, TIKO uses a point-and-grasp strategy. The user places their hand in front of an object and pulses their muscle once for grasp mode or twice for tool mode (illustrated by the headphone example). A second pulse near the object triggers closure, and a final pulse resets the grasp. Because the pipeline is built on open-vocabulary object detection, users can add and customize grasps indefinitely without losing the dexterity of the underlying mechanical body.

Awards: 1st Place in Biomedical Engineering, 2026 Mercer Science & Engineering Fair; Diamond Award; Blue Ocean Award.

I founded LimbKit, a nonprofit that designs low-cost, accessible prosthetics.

LimbKit starts with the fitting. A prosthetic is only as good as its fit, but people with limb differences in developing countries often can’t reach the clinical support a custom fitting normally requires. Our process replaces that visit with two photos of the limb, from which we derive 15 measurements to generate a custom design.

Apart from the foam padding and Velcro straps, the prosthetic is fully 3D-printed, so anyone with a basic consumer 3D printer can make one. Most of the body is rigid PETG or PLA, with a flexible TPU spine running along its length so it can expand and contract as the user’s limb changes size through the day.

Conventional prosthetics imitate the human hand, but five fingers are hard to control intuitively and handle only a narrow range of real-world tasks. LimbKit instead uses a set of 12 attachments that swap on and off quickly, each built for a specific activity. Users can play the violin, eat with cutlery, or write with a pen without thinking about where the fingers land or which grip to trigger.

This project started as a follow-up to my earlier work on gecko-inspired compliant grippers. With six other students, I built a gripper that swaps between gecko adhesive pads and polyurethane pads depending on what it is picking up. The gecko pads beat silicone on almost every object we tried, so I molded a custom flexible polyurethane instead, which held rough objects far better than gecko could.

A camera and a pair of LEDs feed a 2D FFT that estimates surface roughness, while the same image predicts how much of the pad will actually engage. Those measurements, plus mass and a semantic description of the surface, let a vision-language model search prior grasps and predict the force needed. Across 44 objects tested with both pads, a single roughness threshold picked the lower-force configuration 86.4% of the time. On unseen objects, retrieval-grounded prediction reached 0.205 N mean absolute error, a 91% reduction from a zero-shot baseline. I worked on this with a team at the NJ Governor's School of Engineering and Technology.

In the 2026 FRC season, I co-led the mechanical team — driving robot archetype and design strategy while contributing to robot mechanisms. I enjoyed assembling and maintaining our signature turret, an ambitious system pairing two shooters with full 360° rotation.

As intake design lead, my team rapidly iterated through five competition-ready versions. Each release pushed balls-per-second (BPS) higher, with the final design hitting 25 BPS. Rigidity drove the architecture: the first version, a rigid slapdown intake, was abandoned for weight; the second, a flexible four-bar, folded on impact. As match aggression scaled with level of play, our final sliding intake used a curved rack-and-pinion deployment with 1/2" polycarbonate arms — combining the impact-tolerance of the four-bar with the rigidity of the slapdown. Over spring break, I led a sub-team that tested 28 unique compression geometries to lock in the BPS optimum.

Awards: FRC World Championship Creativity Award.

This project investigates the macro-micro coupling between compliant backings and gecko-inspired adhesives to improve robotic grasping on complex geometries. The resulting paper, Maximizing van der Waals Contact in Gecko-Inspired Grippers for Complex Object Geometries via Compliant Backing, was submitted to RoboSoft 2026. Using a controlled 2×2 design, I evaluated gecko adhesives versus silicone pads on rigid (PLA) and compliant (TPU) backings to isolate the effect of macro-conformity on adhesion. The compliant Gecko+TPU configuration significantly reduced required normal preload on smooth, curved objects by maximizing real van der Waals contact area. I also developed and validated a reduced-order model predicting a critical performance crossover, establishing clear design guidelines: compliant gecko grippers are optimal for smooth, curved surfaces, while friction-based methods remain superior for rougher ones.

As scoring mechanism subteam design lead, I led development of a dual-purpose end effector for handling both CORAL and ALGAE in the 2025 FRC game, requiring a single compact mechanism that could switch tasks without added fragility or cycle-time cost. A belt-driven, four-roller architecture with guiding side flanges centers and extracts CORAL from the intake, then stows and actively ejects it for scoring, with real-time sensor feedback to confirm capture and prevent jams. For ALGAE, a paired compliant-wheel "claw" pivots into the reef, rolls the piece deeper into the grip, and halts on sensor detection; reversing the wheels enables fast, controlled ejection for barge scoring. Stacking the ALGAE system above the CORAL path and tuning geometry (~35° CORAL angle) optimized scoring alignment while disciplining mass and packaging. The design converged after 18 prototypes, followed by four refined competition builds balancing reliability, compactness, and match durability.

Awards: FIRST Robotics World Championship — Milstein Division Winner & Einstein Field Qualifier (2025).

This project, my first robotic hand, was intentionally designed for easy assembly with basic tools. I used it as the foundation of a three-day teaching program in Gansu, where students progressed through a structured curriculum introducing core mechanism-design principles, beginning with simple pendant design. Students then assembled the hand from provided parts, learned foundational wiring through basic LED circuits, and integrated the electronics. The program concluded with programming the hand for basic movements and rock-paper-scissors competitions — giving students a tangible way to validate their mechanical, electrical, and software work.