02 · Projects
A model you can only write down is half a model. These are systems I study — cities, causality, macro policy, markets — rebuilt as things a person can actually operate, break, and learn from.
Bilingual neighborhood intelligence reports as a live web product. Fifteen analytical modules across four packages — a "detective's case file" for any address in China: demographics, built environment, commerce, risk. Data pulled at query time, not pasted on a dashboard. My flagship build, and the backbone of my planning-school portfolio.
A city-management simulation where the scarce resource is not money but credibility. You govern with causal inference: policies are interventions, citizens are an interrupted time series, and every election asks you to prove what your term actually caused. v5.3 — archival-paper 2.5D city, municipal hall, bond market, era chains.
A macroeconomic sandbox: you run a small open economy and watch GDP, inflation, liquidity and credit answer back. Ships with a simulation engine covered by a 22-assertion regression suite and in-game concept cards that name the mechanism behind every move. Built to make textbook dynamics falsifiable by play.
A personal equity factor research system, built in two stages: first replicate published broker research factor reports end-to-end (data → construction → backtest → report), then graduate the pipeline to Qlib. Preview reports and a research topic menu are already running; the long-run goal is a fully reproducible, written-by-one-undergrad sell-side-grade process.
The bench is never empty: further builds are planned at the intersection of policy, environment and finance — the same rule applies: real mechanisms, playable, falsifiable.