Undergraduate · Gaoling School of Artificial Intelligence
Renmin University of China
AI Systems ✧
Agent Research ✧
Research Tooling
About / Selected Work / How I Work / Toolkit
I'm an undergraduate at the Gaoling School of Artificial Intelligence, Renmin University of China (RUC), working at the intersection of computer science foundations, AI systems, and practical engineering.
I build tools around real needs and explore how people can work effectively with AI agents. My interests center on human–agent collaboration, agent workflows, and research infrastructure: making execution understandable, evidence inspectable, and decisions revisable as a project evolves.
Alongside these projects, I continue to deepen my grounding in algorithms, systems, machine learning, and mathematical modeling.
PUBLIC ENGINEERING / MATHEMATICAL MODELING / RESEARCH EXPLORATION
|
A campus need, engineered into a usable tool. A searchable archive and companion toolkit for the RUC YOJ online judge, connecting problem data, solution code, verification, and quick-submission workflows. An engineering project spanning data organization, user-facing tools, and automated publication.
|
|
From energy constraints to testable scheduling decisions. A team project on microgrid scheduling for CUMCM 2026, Problem C, covering day-ahead optimization, forecast uncertainty, and intraday rolling decisions. Public materials connect Python implementations with an archived paper, five result workbooks, and explicit validation records.
🏆 First Prize, Beijing Region · CUMCM 2026 · 北京市一等奖 View the repository ↗ · Read the paper ↗ · Browse the release ↗ |
|
Shared understanding between people and agents. A research workspace exploring human–agent collaboration, execution transparency, and evaluation reliability. Connects concrete questions with critical reading, empirical checks, and traceable decisions—without treating an implementation as proof of a research claim.
🔒 Private research · High-level overview only |
|
Research decisions that can evolve with the evidence. Agent-oriented tooling for organizing evidence, revisiting research directions, and adapting plans to changing resources. Combines flexible model-based reasoning with structured project state, versioned records, and explicit handoffs.
🔒 Private tooling · High-level overview only |
Problem first. Start with a real need, define the constraints, and turn an open-ended idea into a concrete question or useful tool.
Evidence before confidence. Keep tests, sources, and reproducible steps close to the claims they support. Distinguish a working implementation from a validated conclusion.
Agent-assisted, human-directed. Automate repeatable work while keeping requirements, review, and decisions explicit. Change the plan when the evidence changes.
AI & experimentation: PyTorch, local model workflows, Ollama, and experiment scripting.
Engineering: command-line tools, automated tests, code review, and reproducible GitHub workflows.
Environment: VS Code · macOS · Linux.
📚 Foundations, coursework & mathematical modeling
Systems & algorithms — C/C++ programming, bit-level reasoning, Data Lab exercises (course fork), and data-structure implementations.
Mathematical modeling — CUMCM 2026 · Problem C — problem formulation, numerical computation, validation, and technical writing for a team competition project. A sanitized public edition includes code, the archived paper, numerical results, and reproduction instructions; official inputs must be obtained separately.
Retrieval & RAG — course assignment — lexical retrieval, vector search, semantic reranking, and retrieval-augmented generation.
Understand the fundamentals. Build something useful. Keep the evidence close.
✦ CURIOSITY / CLARITY / CRAFT ✦