RRocky
中文

AI Agent Engineer · Shanghai / Hong Kong

Turning ambiguous AI ideas into reliable product systems.

I work across AI agents, RAG, LLM evaluation, and Web3 security—connecting research, engineering, and product judgment. This is where I share projects, technical notes, and evolving ideas.

ProjectsGet in touch

02 / WORK

Selected projects

Practical work across generative AI, agent workflows, and security analysis.

View all projects

Founder / Technical Lead

Smart Agent Detector

An explainable AI security platform that turns smart-contract code into evidence-backed risk verdicts.

  • Designed the analysis flow from ABI and normalized AST/IR through multi-agent validation
  • Connected findings, protocol context, and verdicts through a canonical evidence model
  • Coordinated end-to-end delivery across product, backend, frontend, and agent engineering
LangGraphFastAPIReactRAGAST / IRMLflow
GitHub

Data Scientist

Legal Document Agent Workflow

Agent modules for legal content generation, uploaded-document handling, and quality evaluation.

  • Contributed across the full path from requirement discovery to production delivery
  • Improved output quality through retrieval-funnel analysis, prompt refinement, and structured evaluation
  • Turned evaluation and root-cause analysis into a reusable workflow
LangGraphDSPyRAGLLM evaluationAPI integration

AI Project Engineer

LLM-based XML Code Generator

A production-oriented code-generation tool spanning dataset preparation, fine-tuning, prompt refinement, and evaluation.

  • Worked across training-data preparation, fine-tuning experiments, and generation-quality analysis
  • Built repeatable evaluation around accuracy and first-line correctness
  • Helped move the system from an experiment into an internally adopted tool
QwenLoRASFTPrompt engineeringBLEU

03 / NOTES

Recent notes

Writing on AI engineering, product building, and problem solving.

View all notes