Computer Science student interested in artificial intelligence, programming languages, systems software, and computer architecture.

資工系學生，關注人工智慧、程式語言、系統軟體與計算機架構。這個 GitHub Profile 集中展示可執行、可驗證，並清楚標示專案背景與個人貢獻的作品。

Deterministic 15x15 Gomoku engine with Negamax, Alpha-Beta pruning, PVS, transposition tables, tactical threat detection, and time-bounded search.

• Language: Python
• Focus: adversarial search, heuristic evaluation, runtime control
• Evidence: automated rule/protocol tests and passing GitHub Actions

C++ Scheme-like interpreter with a hand-written scanner, recursive parser, S-expression model, lexical environments, closures, special forms, primitives, and structured errors.

• Language: C++17
• Focus: language implementation, parsing, evaluation, lexical scoping
• Evidence: smoke tests pass with GCC, Clang, and MSVC in GitHub Actions

Experimental comparison of sequential partitioning, Linux processes with shared memory, C++ threads, and an in-place QuickSort baseline.

• Language: C++17
• Focus: processes, threads, IPC, performance methodology
• Evidence: reproducible output validation passes on Linux and Windows in GitHub Actions

Five-stage educational Verilog processor with explicit pipeline registers, control/ALU/memory modules, divider and HI/LO paths, and a self-checking Icarus Verilog integration test.

• Language: Verilog
• Focus: pipelined datapath, processor control, simulation, verification boundaries
• Context: four-person team project; personal contribution and publication consent documented

ARMv7 fixed-point Julia set renderer with an AAPCS-correct assembly implementation, cross-compilation workflow, and deterministic 640 × 480 output verification under QEMU.

• Language: ARM assembly and C
• Focus: fixed-point arithmetic, register allocation, ABI correctness, cross-platform verification
• Context: two-person team project; personal contribution and publication consent documented

An nnU-Net v2 3D full-resolution pipeline for cardiac CT segmentation.

• Role: team lead in a three-student team with one faculty advisor
• Result: Public 35/253 (top 13.8%); Private 55/253 (top 21.7%)
• Focus: medical-image segmentation, reproducible experiment configuration, data-governance boundaries

Mamba-inspired multivariate time-series model for predicting two-axis CNC lathe thermal displacement from sensor histories.

• Role: team lead in a three-student team with one faculty advisor
• Result: advanced to the national final; historical internal validation RMSE 1.269709 μm
• Focus: sequence regression, leakage-aware validation, industrial-data constraints

Expert-annotated 419-track dataset associated with the peer-reviewed LSEL music-classification study.

• Focus: research data, expert annotation, citation metadata, rights and limitation statements
• Evidence: public Kaggle release and peer-reviewed paper

React/FastAPI research prototype for local mood journaling, audio-library management, and explainable music exploration.

• Focus: research-to-prototype translation, local-first data, transparent recommendation rules
• Evidence: automated backend tests, production frontend build, and documented research boundary

• RPG Quest Notification System — Java Observer/Strategy design-pattern exercise with JUnit verification.
• Poker Hand Evaluator — dependency-free C11 evaluator with category, tie-break, parsing, and validation tests passing on Linux and Windows.

• Each substantial project has its own repository, build instructions, tests, and limitations.
• Course handouts and instructor-authored reference implementations are not republished.
• Team projects explicitly distinguish personal contribution from team output.
• Competition repositories exclude organizer datasets and distinguish official leaderboard results from internal validation.
• Historical benchmark results are labeled as historical rather than presented as current universal claims.

For application-related contact, please use the information provided in the résumé.