Agentic Systems
Studying self-evolving agents and recursive self-improvement, with a focus on agent harness evolution, automated post-training, and multi-agent systems.
About Me
I am a fourth-year PhD student in the Department of Engineering at the University of Cambridge and a member of Churchill College, supervised by Dr. Stephan Goetz. My research focuses on Natural Language Processing (NLP), Large Language Models (LLMs), and agentic systems. If you are interested in related topics, please feel free to contact me!
Prior to this, I obtained my master’s degree in the Department of Computer Science and Technology at the University of Cambridge. My thesis focused on language models and federated learning, supervised by Dr. Nicholas Lane. I received my bachelor’s degree in Computer Science with Artificial Intelligence at the University of Nottingham.
Education
2023.10 — 2027.09MPhil in Advanced Computer Science
Grade: Distinction.
Supervised by Nicholas Lane.
2022.10 — 2023.09
BSc in Computer Science with Artificial Intelligence
Grade: First-Class Honours.2018.09 — 2022.09
Experience
Honors & Awards
- 2024Master's Prize
- 2022Excellent Graduate
- 2020Dean's Scholarship
- 2019Dream Scholarship
- 2019Dean's Scholarship
Community Service
Invited Talk: Flower Summit 2022 — The C++ ecosystem for Flower, an open-source federated learning framework.
Reviewer: ACL Rolling Review, NeurIPS, ICLR, Scientometrics, EAAI, and IPM Journal.
Research
My research focuses on agentic systems, reasoning & post-training, and LLM generation and evaluation. I welcome collaborations on these and related topics.
Reasoning & Post-Training
Improving the reasoning capabilities of language models through chain-of-thought, data synthesis, supervised fine-tuning, and reinforcement learning with verifiable rewards (RLVR).
LLM Generation & Evaluation
Developing reliable methods for LLM generation and evaluation, particularly for complex and domain-specific tasks.
Publications
- 1.
Benchmarking Patent Drafting from Inventor-Style Disclosures
Conference 2026EMNLP 2026
A benchmark for evaluating patent drafting from realistic inventor-style disclosures.
- 2.
ADIAS: Automated Design of Interactive Agentic Systems
Preprint 2026arXiv:2608.06410
An automated framework for designing interactive agentic systems.
- 3.
Reasoning for Hierarchical Text Classification: The Case of Patents
Conference 2026Findings of ACL 2026
A reasoning-based approach to hierarchical text classification in the patent domain.
- 4.
Enriching Patent Claim Generation with European Patent Dataset
Conference 2025Findings of EMNLP 2025
A European patent dataset for enriching and improving patent claim generation.
- 5.
Towards Better Evaluation for Generated Patent Claims
Conference 2025ACL 2025
A systematic framework for evaluating the quality of generated patent claims.
- 6.
Natural Language Processing in the Patent Domain: A Survey
Journal 2025Artificial Intelligence Review · Q1 · IF 18.8
A comprehensive survey of natural language processing methods and applications in the patent domain.
- 7.
Patent-CR: A Dataset for Patent Claim Revision
Conference 2025NAACL 2025
A dataset designed to support research on patent claim revision.
- 8.
Can Large Language Models Generate High-quality Patent Claims?
Conference 2025Findings of NAACL 2025
An empirical study of large language models for generating high-quality patent claims.
- 9.
FDAPT: Federated Domain-adaptive Pre-training for Language Models
Workshop 2023NeurIPS 2023 FL@FM
Federated domain-adaptive pre-training for adapting language models while keeping distributed data decentralized.
No publications in this category yet.
* Equal contribution.
Citation counts from Google Scholar, updated 2026-09-21.