Yuqin Bai

Master's

Data Science

Leyan Technologies

Software Development Intern

• Reduced clustering time by 75% for better real-time recognition by porting the core clustering module to Python; Optimized query-processing and semantic recognition for better understanding the customer pain points. • Trained models using SimCSE algorithm to generate embeddings on business data with different tokenizing methods and TF-iDF; Improved recall from 0.2 to 0.65 compared to the original Word2Vec embeddings. • Summarized additional positive/counter examples and developed a test set for sentence intent recognition to determine the best intent recognition rules; Reported detailed testing documentation to the team of 7 members.