CV
Education, research, industry experience, and selected publications. Use the PDF icon for the full resume.
Contact Information
| Name | Zheyuan Xiao |
| Professional Title | M.S. Candidate in Computer Science | Software Engineer |
| xiaozheyuan72@gmail.com | |
| Location | Auckland, New Zealand |
| Website | https://zxia545.github.io |
Professional Summary
M.S. candidate in Computer Science at The University of Texas at Austin. My research focuses on large language models, especially evaluation, post-training, data-centric learning, and agentic systems. I am broadly interested in how to make LLMs more reliable and effective through better evaluation targets, better training data, and better system design. My recent work has explored language model evaluation, knowledge transfer in fine-tuned models, and agent-based systems for structured information discovery and decision workflows. Alongside research, I work as a Software Engineer at Resideo in Auckland, New Zealand, where I develop automated verification and validation systems for firmware-based products. Working on reliability, testing, and end-to-end behavior in production environments has shaped my broader interest in building systems that are not only capable, but also measurable, dependable, and useful in practice.
Education
-
2023 - Present Austin, Texas, USA
Master of Science
The University of Texas at Austin
Computer Science
- GPA: 4.0/4.0
-
2018 - 2021 Auckland, New Zealand
Bachelor of Engineering (Honours), First Class Honours
The University of Auckland
Software Engineering
- GPA: 7.029/9
Research Experience
-
2025 - Present Remote
Research Collaborator, TalentScope Copilot
Microsoft Research Asia
Collaboration on an LLM-driven autonomous web agent platform for academic talent discovery.
- Developed a browser-based web agent that searches, navigates, collects evidence, and synthesizes candidate profiles.
- Built multi-agent orchestration with AutoGen and Playwright-based tools for browsing, page interaction, and evidence-grounded extraction.
- Implemented workflow execution, provenance tracking, and evaluation across a TypeScript front end and Python back end.
-
2024 - 2025 Remote
Research Collaborator, LLM Knowledge Inheritance Study
Microsoft Research Asia
Collaboration on how domain knowledge is retained and transferred during LLM fine-tuning.
- Built the end-to-end data and training pipeline for domain-specific QA datasets across multiple knowledge areas.
- Fine-tuned Qwen2.5, LLaMA 3.1, and Mistral 7B models with LoRA and supervised fine-tuning on multi-node 8xA100 clusters.
- Resulting work was accepted at NeurIPS 2025.
-
2024 - Present Remote
Research Assistant
HKUST(GZ)
Collaboration on LLM-based preference evaluation and the reliability of model judgments.
- Investigated desirability and information mass as two central factors affecting evaluation reliability.
- Conducted empirical analysis and synthesized findings into work later published in EMNLP 2025.
Industry Experience
-
2025 - Present Auckland, New Zealand
Software Engineer II
Resideo (Honeywell Home)
- Built an end-to-end automation test system spanning GitHub Actions CI, cloud camera platform integration, and MQTT-based device validation.
- Designed a scalable reliability and reporting architecture for NVR and camera ecosystems across firmware and cloud services.
- Fine-tuned Qwen-Image with LoRA to generate synthetic data for camera analytics and scene understanding.
- Collaborated on CLIP and CNN deployment workflows for Ambarella CV72S smart camera inference.
-
2022 - 2025 Auckland, New Zealand
Intermediate Software Engineer
Fisher & Paykel Appliances
- Led the creation of an automated testing platform integrating hardware, firmware, and higher-level software.
- Developed embedded software in C for refrigerator systems and improved core control algorithms.
- Managed Docker, Makefile, Azure DevOps, and Jenkins-based CI/CD workflows for team delivery.
-
2021 - 2021 Auckland, New Zealand
Teaching Assistant
The University of Auckland
- Assisted with Advanced Topics in Database Systems.
- Assisted with Object-Oriented Software Development.
- Assisted with Functional Programming and Distributed Services.
-
2019 - 2020 Shenzhen, China
Machine Learning Intern
Intellifusion
- Developed and optimized face recognition models for better accuracy and performance.
- Helped integrate face recognition technology into security and surveillance products.
Publications
-
2025 Unveiling the Learning Mind of Language Models: A Cognitive Framework and Empirical Study
NeurIPS 2025
Empirical study of knowledge inheritance during domain-specific LLM fine-tuning.
-
2025 Rethinking LLM-based Preference Evaluation
EMNLP 2025
Study on the roles of desirability and information mass in preference-based evaluation.
Awards
-
2020 First in Course Award in Computer Graphics and Image Processing
The University of Auckland
Recognized for the top course performance in semester one of 2020.