Meet Beichen Hu, a U of I sophomore majoring in Sociology and Statistics. He first learned about the opportunity through the department emails and newsletters. Beichen wish to synthesize the knowledge and skills he acquired in his majors, enabling him to utilize technical skills to tangibly resolve the social problems, so he joined ATLAS internship program. He is currently completing his first semester as a AI Research intern with the ATLAS AI Research Group.
As part of the ATLAS AI Research Group, Beichen is mainly working on developing the "LAS Advising RAG," a specialized conversational AI agent built with Python and LangChain, designing a "Golden Dataset" in JSON format to rigorously and quantitatively evaluate Large Language Models (LLMs) on their accuracy, and conducting a comparative architectural study to evaluate the performance between “Full Knowledge Base Prompting" and "Traditional Chunked RAG" to optimize AI's reliability for all students at U of I.
Through his position, Beichen sharpens his technical skills in building Retrieval-Augmented Generation (RAG) pipelines using LangChain, managing vector databases (ChromaDB), and developing user-friendly frontend interfaces using Streamlit. He also refined skills in data pre-processing and MLOps. Besides technical skills, Beichen also gained communication, problem-solving, and project management skills through his internship.
One of the biggest takeaways from his internship is that an effective AI must not merely be more accurate; it must also be designed from a societal perspective. Only by transforming inherently implicit and complex rules into simple, user-friendly tools can we truly enable more people to access information equitably.
Down the line after graduation, Beichen plans to become a Data Scientist or AI Researcher working at the intersection of Social Science and Artificial Intelligence. When he has time, he spends time on outdoor activities, such as Ince climbing, or outdoor photography.