Nan Tang

nan2.jpg

Associate Prof. and PG Coordinator: Data Science and Analytics Thrust
Associate Dean (PG Affairs): Information Hub
University Senate Member: HKUST(Guangzhou)

I also hold an affiliated position at Hong Kong University of Science and Technology, the Clear Water Bay campus at Hong Kong. Before joining HKUST(GZ), I worked as a senior scientist at Qatar Computing Research Institute, a visiting scientist at MIT CSAIL, a research fellow at University of Edinburgh, a scientific staff member at CWI (national research institute for mathematics and computer science in the Netherlands), and a visiting scholar at University of Waterloo.

I am directing the Data Intelligence and Analytics Lab (DIAL). I am currently focusing on the following projects:

  • Document-to-Database: Unlocking Structured Insights from Documents. Transform unstructured documents (PDFs, reports, emails) into structured database tables so that analytics, joins, and queries operate directly over derived schemas—bridging document-centric content and tabular analytics at scale.
  • Document AI: From Pages to Purpose-Built Intelligence. Leverage AI (vision, NLP, layout, semantics) to ingest, understand and operationalize documents in business workflows—e.g., extracting structured entities, linking documents, automating decision-flows—so documents become active assets, not inert files.
  • Agent Memory for Data-Analytic Tasks: Enabling Persistent Reasoning. Build memory systems for data-analytic agents so they accumulate experiences, recall prior analyses and reasoning chains, and maintain context across multi-step workflows—enabling agents to “remember” prior tables, joins, corrections, and meta-decisions..
  • DeepFund: Real-Time Multi-Agent LLM in Finance. The DeepFund platform by Paradoox AI brings together live market data, multi-agent LLMs (analysts & portfolio managers), and real-time evaluation—moving beyond back-testing to assess LLMs in real investment scenarios with built-in risk/control semantics.

Office: E3 601
E-mail: nantang (at) hkust-gz.edu.cn
Call: (+86)-20-88330888

DBLP Google Scholar

news

Jun 16, 2025 :pencil: [VLDB 2025] Three papers “Weak-to-Strong Prompts with Lightweight-to-Powerful LLMs for High-Accuracy, Low-Cost, and Explainable Data Transformation”, “Data Imputation with Limited Data Redundancy Using Data Lakes”, “AutoPrep: Natural Language Question-Aware Data Preparation with a Multi-Agent Framework”, and a tutorial “Natural Language to SQL: State of the Art and Open Problems” were accepted.
May 16, 2025 :pencil: [KDD 2025] Paper “NL2SQL-BUGs: A Benchmark for Detecting Semantic Errors in NL2SQL Translation” was accepted by KDD 2025 (Datasets and Benchmarks Track).
May 1, 2025 :pencil: [ICML 2025] Paper “Alpha-SQL: Zero-Shot Text-to-SQL using Monte Carlo Tree Search” was accepted by ICML (poster) 2025.
Apr 30, 2025 :pencil: [IJCAI 2025] Paper “RAMer: Reconstruction-based Adversarial Model for Multi-party Multi-modal Multi-label Emotion Recognition” was accepted by IJCAI 2025.
Apr 2, 2025 :pencil: [AIED 2025] Paper “Automatic Modeling and Analysis of Students’ Problem-Solving Handwriting Trajectories” was accepted by The 26th International Conference on Artificial Intelligence in Education.