Kunzhi Wang

Biography

I am a researcher working on efficient large language models, practical reasoning systems, and deployable AI tools. My current interests center on improving model efficiency without sacrificing quality, especially for post-training, adaptation, and evaluation workflows.

Before this, I worked on machine learning systems and multimodal applications, with a long-term interest in turning research ideas into robust products. I care about clarity in writing, pragmatic engineering, and simple interfaces.

Recent News

Show all

2026.04 This homepage was updated to a cleaner academic layout inspired by the reference site.
2025.12 Started a new project on efficient reasoning workflows for compact language models.
2025.09 Released a technical note on model adaptation and lightweight evaluation pipelines.
2025.05 Presented recent work on practical ML systems and reproducible research tooling.

Publications & Preprints

Designing Compact Multimodal Systems for Practical Use

Kunzhi Wang, Collaborator A, Collaborator B

ArXiv Preprint [Paper] [Code]

Data Quality Signals for Faster Model Iteration

Kunzhi Wang, Collaborator C

Workshop on Reliable Machine Learning [Paper]

Lightweight Evaluation Pipelines for Applied AI

Kunzhi Wang, Collaborator D, Collaborator E

Technical Report [Project]

Internship

Research Intern, Example AI Lab

2025 Summer

Worked on efficient post-training pipelines for compact language models and evaluation tooling for applied research.

Machine Learning Intern, Example Company

2024 Summer

Built internal benchmarking workflows and supported deployment-facing model iteration for multimodal applications.

Services

Reviewer Conference on Applied AI Systems
Reviewer Workshop on Reliable Machine Learning
Teaching Assistant Introduction to Machine Learning