M.S. in Computer Science
University of California, Riverside
- Relevant Courses: LLM, NLP, Deep Learning, Trustworthy AI, GPU Architecture, High Performance Computing, Data Mining.
About
Hi there! I'm Hanhui Su, a master's student in Computer Science at the University of California, Riverside. Before that, I earned my bachelor's degree from Beijing Information Science & Technology University.
Before graduate school, I interned at the Institute of Software, Chinese Academy of Sciences, where I worked on large language model-related projects. That experience gave me a practical entry point into LLMs and shaped my interest in continuing in this field.
Currently, I’m especially interested in AI agents and practical LLM applications, particularly systems that connect models with tools, data, and real-world workflows. What I value most is starting from real needs, using technology to solve practical problems, bridging gaps in existing workflows, and turning model capabilities into tools that people can actually use.
Education
B.E. in Internet of Things Engineering
Beijing Information Science & Technology University
- Ranked in the top 5% of the department.
- Relevant Courses: Data Structures, Operating Systems, Computer Organization, Algorithms, Data Processing, Machine Learning.
Internship
LLM Intern
Institute of Software, Chinese Academy of Sciences
- Processed 30 GB of Chinese eBook data for LLM pretraining with noise filtering, deduplication, and structural normalization.
- Contributed to ArkPDFSFT data generation and validation, Llama 2 output evaluation, and LoRA fine-tuning experiments.
Embedded Software Engineer Intern
Beijing Saishu Technology
- Contributed to a Bluetooth debugging tool and optimized networking through firmware updates.
- Reduced device connection issues by 20% and supported deployment of approximately 50 IoT terminals across multiple sites.
Selected Work
Project
FilingLens: Evidence-Grounded RAG Agent
A financial analysis agent for U.S. public companies, designed to help users understand a company faster. Users can ask natural-language questions about fundamentals, cash flow, valuation, risks, or business models, and receive evidence-backed analytical answers with citations and clear limitations.
Project
MarketTone: Fine-Tuned Qwen for Social Market Sentiment
Fine-tuned an LLM for financial sentiment analysis, converting social media posts and investor comments into structured bullish/bearish signals, risk flags, and related ticker information, with a path toward large-scale real-time market sentiment monitoring.
Skills
LLM
PyTorch, Hugging Face Transformers, PEFT, LoRA/QLoRA, Unsloth, Distillation, vLLM.
Agents
LangGraph, ChromaDB, DuckDB, FastAPI, MCP, tool calling, multi-agent workflows.
Languages / Tools
Python, SQL, C/C++, Linux/Shell, Git, Docker, Pandas, NumPy, Pytest.