I earned my bachelor's degree in Artificial Intelligence from Beijing University of Posts and Telecommunications (BUPT) and am now pursuing graduate studies in Computer Science at The University of Hong Kong (HKU).

I enjoy turning ambitious AI ideas into software systems with explicit boundaries, observable behavior, and evidence-backed quality.

My current interests sit at the intersection of:

• product-first AI agents;
• agent runtimes, traces, replay, and evaluation;
• local-first and human-in-the-loop systems;
• software architecture and domain modeling;
• AI-assisted learning and knowledge work.

I care about more than whether an agent can produce a convincing answer. I want to understand why it behaved that way, how it fails, how it recovers, and what evidence would make its result trustworthy.

Product before abstraction. Evidence before claims.
Human judgment where it matters.

An assessment-driven, local-first learning agent built on an observable Agent Runtime and Eval Harness.

TheGrandQuiz does more than help people consume learning materials. It uses grounded conversations and iterative assessment to expose concepts that only feel understood, then carries those weak points into future review.

It brings together:

• grounded ingestion and evidence-linked learning;
• adaptive assessment and durable learning memory;
• reviewable voice input and human correction;
• event-driven agent execution and recovery;
• trace, replay, evaluation, and release gates.

A personal laboratory for reusable AI-agent workflows.

This repository explores how recurring ways of thinking can become focused, composable skills rather than oversized prompts.

Current workflows cover:

• deep reading and assessment;
• software-system mastery;
• requirements reality checks;
• audience-aware project narratives;
• architecture drift audits.

A visual, local-first desktop timer that makes the passage of time tangible.

HyTicker combines a dot-matrix time display, immersive focus sessions, local statistics, and playful interactions in a minimal desktop experience.

• How assessment can reveal the difference between familiarity and mastery.
• How traces, replay, and evals can make agent behavior reviewable.
• How domain models and architectural intent can survive continuous change.
• How reusable skills can encode judgment without becoming rigid checklists.
• How product needs should shape agent abstractions—not the other way around.

I am always interested in thoughtful conversations about AI agents, learning systems, software architecture, and open-source products.

• Email: CillianHe@gmail.com
• GitHub: @Hyr1sky
• X: @Hyrisky