YW.

Selected projects

Work that connects models, products, mechanisms, and access.

These projects differ in scale, but share a method: define the real question, build something testable, and keep the limits of the evidence visible.

Featured work

Letter World V2.0 results from an automated verification round; synthetic test data, not student outcomes.Automated verification · synthetic test data
01 / productV2.0 live

Letter World

Overview

A focused vocabulary-practice platform. An owner-reported 45-day snapshot records 308 students and 310,692 practice responses.

View project
Figure 2: state-preparation RMSE and lowest-two-state weight across the 63-case grid.
02 / modelIndependent research / ongoing

When Is a Two-State Tunneling Model Reliable?

Overview

An independent computational study of how numerical reliability and approximation error change across a controlled parameter grid.

View project
Yuzhe's frame-mounted shifter installed on a pink mountain bike.
03 / mechanismInstalled prototype

Freeing the Handlebar

Overview

A frame-mounted shifter I designed and tested, now installed on my bike using the bottle-cage mounting points to move the control off the handlebar.

View project
AI-generated illustration of a lemon-battery experiment with metal electrodes, connecting wires and an LED; not a photograph of a completed project prototype.Lemon battery · AI-generated illustration
04 / communityIn progress

Rural Physics Initiative

Overview

An in-progress effort to design low-cost, engaging physics experiments for schools with limited laboratory access.

View project

Additional work

Other ways of working through a problem.

Builder

VEX Robotics

Built and revised a competition robot for ball collection based on task requirements and driver feedback. Team 1019B received First Prize and a Design Award in VEX AI at the 2025 Asia-Pacific event.

Related course project

Qiskit Two-State Simulation

Trotterization, an exact two-state evolution benchmark, RMSE, and circuit-depth tradeoffs. This course project concerns digital-simulation error inside an effective model; it is separate from the independent double-well study.