Liuxin Yang

Master's

Electronics Technology

LinkedIn

Software Engineer Machine Learning Intern

- Incorporated raw profile and ads texts and retained key tabular features, reducing derived features and manual engineering.Docker, Git, Spark, Generative AI, LLM, Hadoop, Scala, Python, Airflow Sunnyvale, USA - Designed trait-based prompts for LLM to detect fraud ads accounts, increasing label coverage by 20% at 80% precision - Trained XGBoost model by adding GPT-generated labels with system labels, improving precision by 30% over current system - Built DSPy extensions and adapters, fully automating LLM prompt generation and optimization without manual prompting - Built an OpenConnect pipeline with Docker to automate data processing, LLM labeling, evaluation and review workflows