• Worked on LLM agents and agentic memory, hosted by Dr. Jiaxin Zhang.
• Contributed to RoMem, a memory-oriented method for long-lived LLM agents, focusing on temporal memory representation, retrieval and evaluation.

• Worked on spectral editing and research-assistant RAG systems, hosted by Dr. Keshuang Li.
• Contributed to SEKA, an activation-steering method for LLM alignment and controllable generation, accepted at ICLR 2026.
• Developed a RAG-based research assistant for literature search, technical synthesis and research workflow support.
• The system was actively used by researchers across Huawei labs in the UK, Germany and Italy.

• Worked on synthetic data generation for finance and economics, hosted by Prof. Carsten Maple.
• Contributed to SynthRank, a learning-to-rank approach for generating individual financial transaction data.
• Studied the privacy, utility and fidelity trade-offs of synthetic financial data. The work was presented at AI in Finance for Social Impact at AAAI 2024.
• Collaborated with University College London on time-series financial data generation.

• Supported the GF Bank Z14 mainframe upgrade project, including wiring installation and Hardware Management Console (HMC) switching.
• Joined technical consultation and server-room maintenance projects involving Power 780, S814, DS8870 and DS8880 systems.

• Supported daily maintenance and security upgrades for the internal big-data platform.
• Participated in internal training on the Spark computing framework.

Responsibilities include:

• Deliver tutorials and labs sessions and exam marking.
• Answer questions for all undergraduate and graduate students who choose this course.