Software engineer with 10 years of experience and 3 years focused on crypto exchange full-stack development. Expert in React / TypeScript / Next.js with deep domain knowledge in spot & derivatives trading systems, real-time market data, and multi-chain wallet integration (EVM + Solana). AI Agent Focus: Independently designed and built an AI Tool-Use open-source framework (Python) whose core architecture predated Anthropic's MCP Protocol — enabling LLM-driven autonomous "perceive-decide-act" workflows. Hands-on experience in Tool Schema Design, Function Calling, Multi-Agent Orchestration, RAG Pipeline, Agent Memory Management, and Prompt Engineering. Proficient with LangChain / LangGraph ecosystem. Daily power-user of Cursor / Claude Code for AI-Native development. Former Senior Engineer at Huawei Consumer Business Group, serving 10M+ DAU e-commerce systems.
BYDFI Exchange
Trading System Frontend: Built the frontend for a centralized exchange (CEX) spot and derivatives trading platform with React.js + TypeScript + Vite, supporting real-time quotes, multi-pair comparison, position management, and data visualization with full responsive design across PC / iPad / Mobile. TradingView Integration: Deep integration of TradingView charting library with multi-timeframe switching, custom technical indicator overlays, and interactive chart annotations for professional-grade trading analysis. Real-time Data System: Built a real-time market data Dashboard via WebSocket + REST API — live Order Book, depth charts, and trade history with <100ms update latency. This streaming data expertise directly transfers to LLM Streaming Response architectures. Multi-chain Wallet: Integrated Ethers.js and Web3.js for EVM chains (ETH / BSC / Polygon) and Solana wallet connections (MetaMask, OKX, Phantom), supporting transaction signing, on-chain balance queries, and live data display. AI-Native Engineering Workflow: Designed and deployed an AI-driven development workflow for the team — built a custom Prompt Engineering template library (15+ scenarios: component generation, code review, bug analysis, test case generation), integrated Cursor + Claude Code for an Agent-Assisted code generation and review pipeline. Results: 60% faster repetitive UI development, 40% higher bug detection in code review, 40% overall team delivery speed improvement. AI Agent Trading Prototype: Designed an LLM-based market data analysis Agent with Function Calling integration to real-time market APIs and on-chain data, enabling natural-language-driven market insight queries and trade signal analysis. Performance: Implemented code splitting, lazy loading, image compression, and static API caching — 45% first-paint reduction, Lighthouse score 90+.
Huawei Technologies
Led frontend architecture and development for Huawei Mall core systems — homepage, checkout, payment, cart, and after-sales modules serving 10M+ DAU. Spearheaded department-wide component library: abstracted 50+ reusable business components deployed via Huawei Cloud, adopted by 10+ product lines, boosting team dev efficiency by 35%. Introduced automated testing framework (unit + E2E), achieving 80% code coverage and reducing production bugs by 60%. Authored 80 technical articles on performance optimization, debugging, and architecture best practices — became the core knowledge-base contributor for the division.
Shenzhen Tianyuan DK Information Technology
Independently delivered 3 large-scale enterprise web applications; led migration from jQuery to Vue / React modern frameworks, improving development efficiency by 50%. Rebuilt mobile H5 build pipeline — 40% reduction in bundle size with significant page load improvements.
Bachelor's Degree
Python | AI Agent | LangChain | LLM API | Function Calling | RAG | Prompt Engineering
Independently designed and built an AI Agent open-source framework whose core architecture predated Anthropic's MCP Protocol — enabling LLMs to dynamically discover, select, and invoke external tools (filesystem, browser, database, third-party APIs) for end-to-end autonomous task execution. Agent Architecture: Defined a unified Tool Description Schema and Function Calling Interface with hot-pluggable tool registration and dynamic routing; implemented multi-turn dialogue state management (Agent Memory) and context window optimization for consistent reasoning across long task chains. Tool Chain Orchestration: Built an orchestration engine supporting serial dependencies and parallel execution between tools, enabling automatic task decomposition and scheduling aligned with Multi-Agent Orchestration principles. LLM Integration: Implemented Function Calling and Streaming Response via OpenAI / Anthropic Claude APIs; designed System Prompt templates and CoT (Chain-of-Thought) reasoning chains; integrated an Evaluation Pipeline for quantitative output quality assessment. RAG Enhancement: Built a RAG (Retrieval-Augmented Generation) Pipeline with Vector Database for semantic retrieval and context injection from private knowledge bases, enabling precise domain-specific Q&A and context-augmented reasoning. Project gained traction in the open-source community and validated the architecture; later transitioned to MCP standard protocol-based Agent development, with ongoing iteration.
Plasmo | React | TypeScript | Ethers.js | LLM API | Prompt Engineering | Agent
Built a Web3 browser extension integrating an AI-powered smart contract security Agent — performs real-time security auditing on DApp contract code, automatically detecting reentrancy attacks, permission vulnerabilities, and malicious approvals with actionable risk ratings. Context Injection Strategy: Injects contract ABI, transaction parameters, on-chain state, and historical interaction data into the LLM with domain-specialized System Prompts for precise security analysis, significantly improving risk detection rates. Multi-chain wallet connectivity (MetaMask / OKX / Phantom), real-time Gas monitoring, covering EVM + Solana dual ecosystem.
LangChain / LangGraph | CrewAI | Python | React | TypeScript | MCP
Building a LangGraph-based Multi-Agent on-chain analytics system: Data Collection Agent + Market Analysis Agent + Risk Assessment Agent + Report Generation Agent, orchestrated via Coordinator Pattern for multi-agent collaboration. Python + FastAPI Agent backend with React + TypeScript real-time frontend; supports natural language-driven on-chain data queries (e.g. "Analyze ETH/USDT 24h large-order distribution"). Integrates MCP Server as a standardized interface layer between Agents and DeFi data sources (DEX liquidity, Gas fees, on-chain transactions).
Chinese (Native), English (Professional working proficiency — fluent in technical documentation and daily communication, experienced in remote async collaboration)
80+ technical blog posts covering frontend architecture, performance optimization, Web3 development, and AI Agent development
Actively following AI Agent ecosystem evolution (LangChain / CrewAI / MCP) and DeFi protocol design, on-chain data analytics
Experienced in remote-first, async communication, and cross-timezone team collaboration
Daily power-user of Cursor / Claude Code / GitHub Copilot, continuously exploring AI-Native development best practices