Zoro

My work has long focused on natural language processing, knowledge graphs and LLM agents, with an equal interest in turning models, data and interfaces into systems that actually run.

The experience spans Chinese information extraction, text classification and sentiment analysis, knowledge-graph construction, multi-source data fusion, RAG, tool use and automated workflows. Instead of listing employers and titles, this page keeps the method: understand the problem, organise the data, build the model, connect the system, then verify the result with tests and feedback.

ZoroTreeking is a long-running record beside that work. It brings together AI learning, market observations, photography and hiking — how technology works, and how people make sense of change.

• ZoroTreeking

  Public project

  A personal publishing and data-automation lab with bilingual content, AI summaries, market data, photo stories, hiking notes and automated delivery.

• LLM Agents and Workflow Automation

  Recent practice

  Agent workflows for long-document understanding, data analysis and report generation, covering retrieval, tool use, result validation and operational monitoring.

• Knowledge Graphs and Data Fusion

  Engineering practice

  End-to-end work from text extraction, entity and relation modelling, and data cleaning to graph storage, queries and complex-information analysis.

• Chinese NLP Information Extraction

  Algorithm practice

  Event extraction, named-entity recognition, text classification and sentiment analysis, including dataset design, training, evaluation and deployment.

• Productionising Algorithm Systems

  Ongoing

  Connecting model capabilities to backend services, data pipelines and interfaces, with attention to reliability, maintainability and real feedback.

• 📖
  To Live— Yu Hua

  The force of life itself

• 📖
  Ordinary World— Lu Yao

  How ordinary people find their light

• 📖
  Those Things About the Ming Dynasty— Dangnian Mingyue

  History, inevitability and chance

• 📖
  Sword Snow Stride— Fenghuo Xizhuhou

  Character arcs inside a long story

• Move LLM agents from demonstrations into stable, maintainable workflows
• Keep growing from implementing algorithms to defining problems and validating solutions
• Record honest observations across technology, markets, photography and hiking
• Stay data-driven while leaving room for uncertainty

1. Foundations
   Built a foundation across computing, machine learning and data systems.
2. NLP Practice
   Started with Chinese information extraction, text classification and sentiment analysis.
3. Graph Engineering
   Extended extraction into knowledge graphs, data fusion and complex-relation analysis.
4. Agent Engineering
   Moved into LLM applications through RAG, tool use, orchestration, evaluation and monitoring.
5. Continuous Work
   Built ZoroTreeking to keep learning, system practice and real-world notes together.