Yuyang Zhang

Samana Technologies, VIC

Built career-tech product features across assessment flows, trace-data capture, scoring, conversational AI, and low-latency avatar interactions.

• Built a game-based Angular assessment module for a career guidance platform, capturing student choices, timing, hover behaviour, and free-text inputs to support trait profile calculation.
• Implemented interactive micro-simulation flows with responsive Angular components, including card selection, timed decision tasks, drag-and-drop sequencing, and mobile-friendly interaction patterns.
• Contributed across Angular, Node.js/Express, Python/FastAPI, PostgreSQL, and Redis for session orchestration, trace-data capture, scoring, and profile persistence.
• Optimised LangChain and LangGraph workflows for AI-assisted career coaching features, improving orchestration reliability and response flow.
• Integrated and tuned ElevenLabs voice AI APIs, reducing voice-interaction latency by approximately 60% to support near real-time conversation.
• Improved AI avatar interaction latency by approximately 30%, making career-coach conversations feel more responsive and natural for users.

AI Collaboratory, The University of Queensland

Built backend services and database systems to support patient mask-fitting data management, focusing on secure APIs, stability, and scalable processing.

• Built secure RESTful APIs in ASP.NET Core with validation and business logic for patient mask-fitting data management.
• Designed SQL Server data access with EF Core and improved stability by resolving critical bugs and tuning database operations.
• Integrated RabbitMQ for asynchronous message handling, improving backend scalability for large-scale data processing workflows.

AI Collaboratory, The University of Queensland

Conducted research on neural network optimisation, focusing on the analysis of loss landscapes and solution spaces.

• Analysed neural network loss landscapes and solution spaces using PyTorch and Matlab across different architectures.
• Produced reproducible experimental code and documentation to support ongoing AI research into optimisation behaviour.