PE ZH

I'm a native Mandarin Chinese speaker and Computer Science master's student focused on evaluating AI/ML outputs, debugging model behavior, and documenting technical reasoning clearly.

For NeuroGolf / ARC-AGI, I built Python tools to audit AI-generated ONNX models, validating output types, shapes, task semantics, and training-example consistency. I localized subtle model failures through controlled batch evaluation and clear failure documentation.

I built Neko.exe, a FastAPI-based AI agent with contextual response generation, structured JSONL logs, repeatable evaluation hooks, transcript-quality checks, classification, memory controls, dry-run testing, and fallback behavior.

I've also designed an evidence-tracking LLM workflow for research continuity, evaluated motion-analysis pipelines using quantitative and edge-case checks, and contributed merged work to the Cardano Foundation after source validation and maintainer review.