YUYANG ZHENG / SYDNEY & CHINA

Zachary.

AI engineer. Product builder.

Somewhere between logic and wonder.

I build agents, tools, and thoughtful digital experiences — turning experiments into things people can use.在理性与想象之间，持续探索。

View résumé 查看简历 Explore the lab

EXPERIENCE, EDUCATION & SELECTED WORK BELOW

The lab — try what I’m building

04 EXPERIMENTS

LLM Chat Platform

Conversations, tools & model experiments.

Texas Poker AI

Strategy, decision making & game experiments.

LocaLingo

Bilingual reading, offline dictionary & spaced review.

NoteMind AI

Read sources, take anchored notes & practise with AI quizzes.

The person behind the work

Yuyang Zheng
(Zachary)

郑宇扬 · 关于我

Building useful AI.
Understanding how it works.

I build and evaluate LLM agents, RAG systems, and fine-tuned models. My work brings together product engineering, reproducible experiments, and responsible AI — with a focus on tools that support real decisions and everyday knowledge work.

专注 LLM Agent、检索增强生成与模型微调，让 AI 从实验走向真实场景。

GitHub ↗Based in Sydney · Connected to China
01 /

Experience

实践与经历
Jun – Sep 2026Shanghai, ChinaInternship

ByteDance

Backend Engineer
Content Safety · Data-TnS

  • Collaborated with algorithm and MLOps teams on the international Trust & Safety platform, contributing to architecture, code, and ML system tuning for risk detection.
  • Improved ML application integration and helped define code and testing standards to reduce the cost of managing models across business scenarios.
  • Supported high-concurrency, reliability, data-isolation, and system-decoupling work, alongside research into heterogeneous computing.

参与国际内容安全平台与机器学习工程，关注可靠性、系统架构和开发规范。

Trust & SafetyMLOpsBackend systems
Mar – Jun 2026Shanghai, ChinaInternship

Momenta

AI Agent Engineer

  • Developed a Feishu CLI that turns conversational commands into schedules, Lark Project updates, crawler outputs, and business documents.
  • Built asynchronous crawling and data-cleaning pipelines, then generated standardized presentations and multilingual documents from extracted indicators.
  • Built a text-to-SQL analytics agent and dashboard for business-development and operational analysis.

将对话转为实际行动，打通业务自动化、数据分析与文档生成。

Tool-using agentsText-to-SQLAutomation
Dec 2025 – Mar 2026Shanghai, ChinaInternship

Sangfor Technologies

AI Agent Engineer

  • Built multi-agent document generation, question answering, and review workflows using LangChain, LangGraph, FastAPI, Milvus, Redis, and Docker.
  • Implemented hybrid retrieval, intent routing, authentication, rate limiting, and streaming controls for an internal RAG knowledge platform.
  • Created document templates with python-docx, python-pptx, and Jinja2, using chunked generation and review for long business documents.

构建企业知识平台与多 Agent 文档工作流，完善检索、权限和流式体验。

RAGLangGraphDocument automation
Jan – Mar 2024Guangzhou, ChinaInternship

Deloitte

Information Technology Specialist

  • Built Python and REST API automation for ticket synchronization, exception alerts, and routine operational reporting.
  • Integrated OCR document processing for batch recognition, data cleaning, and structured storage.
  • Supported internal IT monitoring, troubleshooting, and user support across engagement teams.

通过流程自动化与 OCR 文档处理，提高内部 IT 工作效率。

PythonREST APIsOCR
02 /

Education

学习轨迹

University of New South Wales

Master of Financial Technology

Postgraduate study in financial technology, AI systems, and data-driven decision making.

新南威尔士大学 · 金融科技硕士
Sep 2026 – Jun 2027Sydney, Australia

University of Tasmania

Bachelor of Information Systems

Coursework and projects in information systems, software engineering, databases, and analytics.

塔斯马尼亚大学 · 信息系统学士
Sep 2023 – Jun 2026Australia

Shanghai Ocean University

Bachelor of Information Management
and Information Systems

Training in management information systems, business processes, and applied programming.

上海海洋大学 · 信息管理与信息系统学士
Oct 2022 – Jun 2026Shanghai, China
03 /

Selected work

项目探索
01Apr 2026

Go Code-Review Agent

A learning and standards platform that connects deterministic code checks with personalized AI feedback.

  • Combined golangci-lint, AST context extraction, LLM reasoning, and sandbox verification, keeping static-analysis findings distinct from generated suggestions.
  • Used diff chunking, semantic caching, and dynamic model routing to manage latency and token cost — from formatting fixes to concurrency and architecture reviews.
  • Designed working, episodic, semantic, and procedural memory with vector retrieval and developer profiles for personalized few-shot feedback.

Research focus / Decision support, benchmark design, and hallucination-risk measurement.

面向 Go 开发者的代码审阅、规范学习与个性化反馈平台。

Go · ASTAgent memorySandbox verification
02Feb 2026

Multi-Agent Litigation Drafting

A document workflow for credit-card overdue cases, with explicit verification and privacy safeguards.

  • Designed an extract–retrieve–draft–verify LangGraph workflow, combining agent planning with deterministic Python checks and identifying claims that require legal review.
  • Built multimodal parsing and a jurisdiction and legal-rule knowledge base, with deterministic checks for amounts, constraints, and interest-rate rules.
  • Prepared hundreds of standard pleadings and fine-tuned Qwen-32B with LoRA to align outputs with document structure and filing formats.
  • Added a privacy gateway that replaces personal information with placeholders before model calls; scoped evaluation to test cases and rule coverage.

Research focus / Fine-tuning, retrieval evaluation, output verification, and privacy-preserving agents.

结合检索、模型微调与规则校验，辅助法律文书起草。

LangGraphQwen · LoRAPrivacy by design
03Dec 2025

Roots & Roads

An education and community platform helping immigrant students aged 12–18 navigate cultural adaptation.

  • Built the full-stack platform with React, TypeScript, Vite, REST APIs, JWT sessions, and WorkOS SSO.
  • Implemented a streaming AI tutor with fallback handling, alongside resource discovery, mood tracking, and Pomodoro task modules.
  • Built a cloud data-cleaning pipeline and documented API contracts, deployment across environments, and automated tests.

Research focus / Full-stack prototyping, reliable streaming inference, and student-facing AI evaluation.

为新移民青少年提供学习、社区支持与文化适应工具。

React · TypeScriptAI tutoringFull-stack
04 /

Tools & thinking

技术与研究

AI & models

LLM workflows · RAG · Vector search · LangChain · LangGraph · LoRA / PEFT · Agent orchestration · Prompt & retrieval evaluation

Engineering

Python · FastAPI · React · TypeScript · SQL / MySQL · Docker · Linux · Git · Redis · CI/CD

Methods

Data cleaning · Web crawling · Retrieval evaluation · Structured error analysis · Benchmark design · Reproducible documentation

What I’m exploring

Business AI, human–AI collaboration, responsible AI evaluation, AI-assisted decision workflows, and document & knowledge-work automation.

Email ZacharyCallyuyangzheng0000@gmail.com · +61 0340778319