Yihao Shu Email and Phone Number

• Applied Parameter Efficient Fine-Tuning (PEFT) using Low-Rank Adaptation (LoRA) and Instruction Fine-Tuning methods to improve the accuracy of the Llama-2-13b Large Language Model for sentiment analysis, achieving an accuracy improvement to 84%.• Implemented and optimized BART and T5 models through advanced prompt engineering to summarize original data texts for sentiment scoring. Employed Zero-Shot Text Classification focusing on Multi-Genre Natural Language Inference (MNLI) to select texts and enhance the correlation of data to target stocks or cryptocurrencies.• Developed a comprehensive sentiment scoring pipeline on AWS SageMaker, automating hourly processes including data acquisition, summarization, classification with weight assignment, sentiment scoring, and transformation into positive/negative expectancy values. These results were stored in AWS S3 and utilized as factors in trading strategies for cryptocurrencies and stocks.