🎓 Year 2 Student @ University of Melbourne - Computer & Software Engineering

💼 AIGC Development Intern

💡 I'm passionate about building AI-powered applications and elegant solutions to complex problems. When I'm not coding, you'll find me playing the cello or exploring new technologies.

• 🔭 Currently working on AI-driven web applications
• 💻 Experienced in React, Next.js, and TypeScript
• 🤖 Interested in AI integration and full-stack development
• 🌱 Always learning and growing as a developer
• ⚡ Fun fact: I change my wallpaper daily for motivation!

• 🔬 AIGC 研发实习生（AI Agent & 全栈） @ Tezign (特赞) · Shanghai · Oct 2025 - Present
  Building AI Agent systems and full-stack features for AIGC products. https://m.tezign.com/home
• 🎓 Computer Science Student @ University of Melbourne - Focusing on algorithms, data structures, and full-stack development
• 📝 Tech Blogger @ joyehuang.me - Sharing knowledge and practical experiences

A full-stack web application that provides voice-driven interview simulation with real-time AI feedback to enhance interview preparation.

Key Features:

• 🎯 Real-time voice recognition and synthesis using Vapi Voice SDK
• 🤖 Dynamic question generation and semantic feedback powered by Google Gemini AI
• 🔐 User authentication and session management with Firebase
• 📊 Personalized review and feedback storage in Firestore
• ⚡ Server-side rendering with Next.js App Router for optimal performance
• 🏗️ Modular architecture for scalability and maintainability

Tech Stack: Next.js · React · Firebase · Gemini AI · Vapi Voice SDK · TypeScript

• 🧭 Designed distributed multi-agent orchestration and communication with Function Calling/Tool Schema, enabling streaming pipelines with queue/event-driven routing, de-duplication, watermarking, resumption, and observability.
• 🧩 Built long-context pipelines (memory tiers, summarization, retrieval stitching, token budget control) with prompt templating and guardrails for consistent outputs.
• 🧬 Developed social-data-driven persona modeling (Profile/Behavior/Motivation) for personalized generation and preference alignment under goal constraints.
• ❓ Implemented counter-questioning for intent clarification to convert ambiguous requirements into structured constraints (goals, audience, style, limits).
• ⚙️ Delivered async tool-calling runtime (checkpoint, retries, idempotency, timeout/rollback, recovery) plus offline replay/evaluation harnesses.
• 🔒 Ensured production-grade delivery with schema validation, modular architecture, auto-generated diagnostic HTML reports, and AWS Marketplace SaaS Contract integration from subscription → permissions/quota → delivery.

• 📚 Curated a structured learning note series for LLM foundations and practical workflows.
• 🧱 Built a VitePress-based one-stop tutorial site for fast navigation and continuous updates.