Yunhua Pei
PhD Candidate in Computer Science
University of Bristol, UK · Senior Research Associate, AI4CI Hub
ge22472 [at] bristol.ac.uk
About
I am a final-year PhD candidate in Computer Science at the University of Bristol, supervised by Prof. John Cartlidge and Dr. Jin Zheng, and a Senior Research Associate at the AI4CI Hub. My thesis, Market, Value, and Momentum, builds a graph-based, multidimensional framework for stock market forecasting.
I build and validate forecasting models for noisy, non-stationary market data. Recent work covers stock forecasting with dynamic graph contrastive learning, multi-agent LLM systems for crypto portfolio allocation, unsupervised anomaly detection in multivariate time series, and evaluating LLMs against market participants' revealed positions and trades.
Before the PhD, I spent nearly two years as a Senior Quantitative Trading Development Engineer in the Financial Markets Division of China CITIC Bank's head office, forecasting Chinese government bond yields for the bond trading desk. During the PhD I also worked with Stratiphy.AI on multimodal market forecasting through an Innovate UK-funded project.
Research interests: trading systems, LLM agents, graph learning, time-series anomaly detection.
I am currently seeking industry and academic opportunities in fintech and LLM agents. Feel free to reach out by email.
News
- 2026Took over AuditOps at the AI4CI Hub, an LLM-agent system that verifies claims against public company filings.
- 2026Two first-author papers accepted as posters at NeurIPS 2026: StakeBench and Market Regime Council.
- 2025Paper on cross-modal temporal fusion for financial market forecasting published at ECAI 2025.
- 2025Received the Best Paper Award at ICAART 2025.
Publications Google Scholar
2026
-
Advances in Neural Information Processing Systems (NeurIPS), 2026 (Poster)
-
Advances in Neural Information Processing Systems (NeurIPS), 2026 (Poster)
2025
-
European Conference on Artificial Intelligence (ECAI), 2025
-
International Conference on Agents and Artificial Intelligence (ICAART), 2025
EARLIER
-
International Conference on Computer Animation and Social Agents (CASA), 2020
-
Image and Graphics Technologies and Applications (IGTA), 2019
Preprints & Journal Extensions
-
Submitted to WSDM 2027
-
Extended journal version of the ICAART 2025 paper
Patent
-
Australian Innovation Patent 2018101512
Experience
Senior Research Associate, AI4CI Hub
10/2026 – present
University of Bristol, UK · Project AuditOps (Innovate UK-funded R&D)
Continuing development of AuditOps, an LLM-agent system that verifies claims against SEC and Companies House filings using retrieval, deterministic calculation and citation-checked extraction.
Senior Research Associate (Quantitative Trading)
06/2024 – 06/2025
Stratiphy.AI, London, UK · seconded via the University of Bristol AI4CI Hub
Built CMTF, a transformer model that fuses FTSE 100 prices, macro indices, news sentiment and financial reports for interpretable price-direction forecasting.
Senior Quantitative Trading Development Engineer
03/2021 – 11/2022
China CITIC Bank (Head Office), Financial Markets Division, Beijing
Built and deployed a multi-factor LSTM strategy for China government bond yields with the bond trading desk, and raised long-term forecast precision by 15.3% by combining model output with technical analysis and market views.
Education
PhD in Computer Science
02/2023 – 02/2027 (expected)
University of Bristol, UK
MSc in Data Science
09/2019 – 11/2020
University of Sheffield, UK
BSc in Information Management and Information Systems
08/2015 – 06/2019
Northwest A&F University, China
Honors & Awards
- Best Paper Award, ICAART 2025
Professional Activities
- Conference reviewer: NeurIPS, AAAI
- Journal reviewer: ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)
Teaching
Graduate Teacher, University of Bristol
- Advanced Financial Technology
- Data Science Mini-Project
- Large-Scale Data Engineering