Jinyuan ZhouJinyuan
Zhou.

Unix Shell Demo written purely in Rust

Built an interactive Unix shell in Rust around a lexer–parser–AST–executor architecture. Implemented quoting, escaping, tilde expansion, built-ins, external programs, pipelines, file-descriptor redirection, command sequencing, &&/|| short-circuit evaluation, relative and absolute path resolution, PATH lookup, and unit/integration tests for parsing and process behavior.

MAY 2026 — PRESENT · Personal project · Developer

• Rust

Reversi AI with alpha–beta search

Ranked 2nd out of 440 students in the APS105 Computer Fundamentals Reversi competition. Built an algorithm to evaluate board positions and choose moves, using minimax, iterative deepening, move ordering and alpha–beta pruning to search deeper within a fixed time limit.

MAR 2026 · University of Toronto APS105 · Developer

• C

Your whole degree in one place: courses, grades, schedule, tasks, and an AI assistant that knows all of it.

Semestra is a semester planner for university students. Set up your program once, add terms and courses, and Semestra keeps the rest connected: the gradebook works out your current mark and what you need on the final, the calendar knows your lectures, labs and exams, deadlines turn into tasks, and Sem, the built-in assistant, can read and update all of it in plain language. This repository is a ground-up rewrite of the original Semestra. Every feature is a module that declares typed capabilities, so the same operation is available in the UI, to the assistant, to plugins and over MCP.

NOV 2025 — PRESENT · Personal project · Developer

• React 19
• TypeScript
• FastAPI
• SQLite
• Pydantic AI

School LMS timetable to calendar

Fetched and processed school LMS data with Python web crawling and networking libraries to generate importable calendars. Optimized parsing for up to 2× speed and over 50% smaller files, so students could import their timetables more easily.

AUG 2024 — JAN 2025 · Nanjing Foreign Language School · Developer

• Python

Self-verification in LLM chain-of-thought

Designed experiments on the relationship between LLM parameter count and repetitive self-verification in chain-of-thought reasoning. Built a multithreaded Python pipeline to query models through OpenAI-compatible APIs, collect reasoning traces and automatically process results across repeated trials. Designed and integrated a semantic classification workflow using Transformer embeddings and logistic regression to identify self-verification statements.

APR 2025 — MAY 2025 · AP Statistics 12 Research Project

• Python