Nan Fung International Properties

Associate - Data Science (Jan 2024 – Present)

● Built an agentic Q&A chatbot for real estate acquisition, achieving 80% accuracy in database retrieval by using RAG-enhanced few-shot prompting to optimize SQL generation with domain knowledge and specialized calculations.

● Developed an adaptive AI prompting framework, improving business team-approved results by 1.75x across multiple use cases.

● Built an LLM evaluation tool, reducing trial and error by 60% to streamline model selection.

● Optimized vector and graph RAG with cross-encoding, boosting retrieval accuracy by 60% for better insights in daily operations.

● Applied CART analysis to marketing campaigns, identifying high-potential customer segments with a 4x increase in conversions, shaping future marketing strategies.

● Led AI training for London & Boston teams (15+ participants/session), upskilling associates to directors.

● Optimized an internal research platform, implementing asynchronous and batch processing to triple request handling speed.

● Presented AI and development updates to the Group CEO, driving strategic discussions.