Jiawei's Picture

Jia-Wei Liao

Ph.D. Candidate
National Taiwan University

About Me

I am a Ph.D. candidate in Computer Science at National Taiwan University and a member of CMLab, advised by Prof. Cheng-Fu Chou. I also collaborate closely with Prof. Tsung-Wei Ke and serve as a Research Assistant at Academia Sinica under the supervision of Dr. Jun-Cheng Chen. My research focuses on generative models, particularly diffusion models and flow matching, with applications in controllable visual generation, inverse problems, and robot manipulation.

Before joining NTU, I received my M.S. degree in Applied Mathematics from National Yang Ming Chiao Tung University, advised by Prof. Wen-Wei Lin. During my master’s studies, I worked on three-dimensional measure-preserving computational geometry and deep learning for medical image analysis. This training established my foundation in applied mathematics, numerical computation, and mathematical modeling.

My journey through internships at Appier, Intel, and SinoPac has strengthened my problem-solving skills, engineering capabilities, and product-oriented mindset. Along the way, I have learned to identify real-world needs, collaborate effectively across disciplines, and build solutions that balance technical feasibility, user value, and business considerations.

Research

I am broadly interested in computer vision and generative AI, with the goal of drawing mathematical and theoretical insights from practical problems and developing algorithms that are controllable, efficient, and reliable. My research focuses on the following three directions:

  1. Visual Perception: Recovering high-quality images from degraded, incomplete, or noisy observations, and enabling machines to build computational representations of the world’s structure and state.
  2. Generative Modeling: Developing diffusion and flow matching models, particularly through inference-time guidance and structured prior modeling, to improve generation quality, controllability, and efficiency.
  3. Trustworthy AI: Investigating the safety of generative models, including adversarial attacks and machine unlearning, to uncover their vulnerabilities and design more reliable models and algorithms.

Education

NTU Logo

National Taiwan University

Ph.D. Candidate, Computer Science and Information Engineering

Sep. 2022 - Present

NYCU Logo

National Yang Ming Chiao Tung University

M.S., Applied Mathematics

Sep. 2020 - Aug. 2022

NTNU Logo

National Taiwan Normal University

B.S., Mathematics

Sep. 2016 - Jun. 2020

Work Experience

Academia Sinica Logo

Academia Sinica

Research Assistant

Jul. 2023 - Present

Intel Logo

Intel

Software Engineer Intern

Jul. 2024 - Aug. 2024

Microsoft Logo

Microsoft

Research Intern

Mar. 2024 - Jun. 2024

Appier Logo

Appier

Data Analyst Intern

Aug. 2022 - Jun. 2023

NCTS Logo

National Center for Theoretical Sciences

Undergraduate Student Researcher

Jul. 2020 - Aug. 2020

Academic Society

GDSC Logo

Member

NTU Google Developer Student Club

Sep. 2024 - Present

NTU Data Analytics Club Logo

5th Director of Academic Affairs

NTU Data Analytics Club

Jun. 2023 - Jun. 2024

TMBA Logo

Member

Department of Strategic Marketing, TMBA

Sep. 2022 - Jun. 2023

Awards & Scholarships

Academic Service