AI Agent Harness Architect · Associate Professor
🎓 Ph.D. in AI & Engineering | B.Eng & M.Eng

I am an AI System Architect and Researcher operating at the intersection of Agentic Retrieval-Augmented Generation (RAG), Multimodal Large Language Models (LLMs), and Knowledge Graphs (KG). I specialize in designing fault-tolerant, budget-aware autonomous agents capable of structuring highly noisy, real-world data (PDFs, DOCX, images) into actionable intelligence.

If it involves parsing the unparsable, extracting the invisible, or orchestrating multi-agent systems to solve domain-specific engineering challenges—I'm building it.

Core: Python, PyTorch, LangChain, Llama, DeepSeek, Claude
Databases/Graph: Neo4j, PostgreSQL, Elasticsearch

• [2026] 🥇 1st Prize (Knowhere AI) - National AI Application Scenario Innovation Challenge, CAAI
• [2026] 🏆 Grand Prize (Snap-Fill) - 1st AI Agent Global Special Competition, CAAI
• [2025] 🏆 Grand Prize (Knowhere AI) - 1st Tsinghua Innovation & Entrepreneurship Competition
• [2025] 🥇 1st Prize (Knowhere AI) - Intel AI Innovation Application Competition
• [2025.08] 🥇 1st Place - Alibaba Tianchi Document End-to-End Structured IE (1/406)

My research focuses on deep learning, ontology, and LLM-driven decision-making in complex environments (e.g., AEC - Architecture, Engineering, and Construction). 66 publications · h-index 25 · 5 ESI highly cited / hot papers.

Selected Core Papers (First/Corresponding Author):

• [Advanced Engineering Informatics 2025] Retrieval augmented generation-driven information retrieval and question answering in construction management (ESI Highly Cited, JCR Q1, CCF-B, IF 11.5)
• [Knowledge-Based Systems 2023] Knowledge graph-enabled adaptive work packaging approach in modular construction (JCR Q1, IF 8.0)
• [Computer-Aided Civil and Infrastructure Engineering 2022] Graph based deep learning model for knowledge base completion in constraint management of construction projects (JCR Q1, IF 9.9)
• [Automation in Construction 2021] Hybrid deep learning model for automating constraint modelling in advanced work packaging (JCR Q1, IF 12.6)
• [Automation in Construction 2020] Ontological knowledge base for concrete bridge rehabilitation project management (ESI Highly Cited, JCR Q1, IF 12.6)

threads:
  - id: 0x01
    status: "SUBMITTED"
    venue: "EMNLP 2026"
    task: "Where vs What, Structure–Content Decomposition: metrics that separate whether a model recalls the correct value from whether it places that value at the correct grid/path coordinate; structure-aware verifiable RL rewards for JSON and table tasks."
  - id: 0x02
    status: "WORKING_PAPER"
    venue: "NAACL 2027"
    task: "Hierarchy as agent harness: graph/tree hierarchy as an explicit navigation harness for long, complex information search; compare retrieval across parsers and SOTA agents."
  - id: 0x03
    status: "WORKING_PAPER"
    venue: "ICLR 2027"
    task: "Mechanistic Study of Hierarchical Representation in Embedding Models: how flat vs tree vs graph representations affect retrieval, and dynamically selecting representation per query/data."