YuntianWang

Deep Learning based Diffractive Optics Optimization

2023 - Present

UCLA ECE, Advisor: Aydogan Ozcan

• Developed a diffractive surface-based pipeline for universal optical waveguide design, which can be streamed down to various tasks such as redirecting, filtering, splitting, and polarization maintaining.
• Designed a diffractive surface system capable of selectively transmitting different classes of images based on the polarization. Bringing innovative methods for privacy protection.

Structure Health Monitoring using Diffractive Processors

2024 - Present

UCLA ECE, Advisor: Aydogan Ozcan

• Designed a hybrid (optical plus digital) system utilizing deep learning to encode vibration spectrum of a building optically and decode by digital network, providing fast and high accuracy structure health monitoring solutions.

Transferable and Universal Adversarial Attacks on Computational Pathology Models

Mar 2025 - Present

UCLA ECE, Advisor: Aydogan Ozcan

• Developed Universal and Transferable Adversarial Perturbations (UTAP) that reveal critical vulnerabilities in pathology foundation models. UTAP comprises a fixed, visually imperceptible microscopic noise pattern that systematically disrupts feature representations, causing significant performance drops across diverse datasets and multiple black-box foundation models.
• Demonstrated both universality (dataset-independent perturbation) and transferability (effective against unseen external models), establishing a benchmark for robustness evaluation and highlighting the need for defense mechanisms in AI-driven pathology and optical microscopy.

Nonlinear Optical Computing with Diffractive Processors

Jan 2026 - Present

UCLA ECE, Advisor: Aydogan Ozcan

• Demonstrated that linear diffractive systems can perform large-scale nonlinear function approximation under incoherent or partially coherent illumination, achieving snapshot computation of up to one million distinct nonlinear functions in a single forward pass.
• Introduced a compact single-layer encoder-decoder co-localization architecture that achieves programmable nonlinear input-output mappings without nonlinear optical materials, proven to be a universal approximator for arbitrary band-limited nonlinear functions.

Optical Diffusion-based Generative Models

Dec 2024 - Jul 2025

UCLA ECE, Advisor: Aydogan Ozcan

• Involved in building an optical generative model capable of generative multiple classes of images including simple dataset to Van Gogh style artwork. The optical generative model drastically decrease the computational power required for novel image generation.

Ph.D. in Electrical and Computer Engineering

Sep 2023 - Present

University of California, Los Angeles

GPA: 3.975/4.0 | Advisor: Aydogan Ozcan

Bachelor in Electrical and Electronics Engineering

Sep 2019 - Jun 2023

Southern University of Science and Technology

GPA: 3.89/4.0 | Advisor: Perry Ping Shum

Amazon AI PhD Fellowship

2025 - 2027

Emil Wolf Outstanding Student Paper Competition

2025

Frontiers in Optics (FiO)

Departmental Fellowship

2023 - 2025

University of California, Los Angeles

Top 10 Undergraduate Student in Engineering School

2023

Southern University of Science and Technology, Shenzhen

Universal and Transferable Attacks on Pathology Foundation Models using Microscopic Perturbations

Jun 2026

Light: Science & Applications

Yuntian Wang, Xilin Yang, Che-Yung Shen, Shuhang Dong, Nir Pillar, Aydogan Ozcan

Breaking the Cascade: Compact Nonlinear Optical Computing with Single-Layer Encoder-Decoder Co-Localization

May 2026

arXiv Preprint

Yuntian Wang, Alexander Chen, Md Sadman Sakib Rahman, Aydogan Ozcan

Large-scale Nonlinear Optical Computing with Incoherent Light via Linear Diffractive Systems

Mar 2026

arXiv Preprint

Alexander Chen, Yuntian Wang, Md Sadman Sakib Rahman, Yuhang Li, Aydogan Ozcan

Structural Vibration Monitoring with Diffractive Optical Processors

Mar 2026

Science Advances

Yuntian Wang, Zafer Yilmaz, Yuhang Li, Edward Liu, Eric Ahlberg, Farid Ghahari, Ertugrul Taciroglu, Aydogan Ozcan

Optical Generative Models

Aug 2025

Nature

Shiqi Chen, Yuhang Li, Yuntian Wang, Hanlong Chen, Aydogan Ozcan

Optimizing Structured Surfaces for Diffractive Waveguides

Jun 2025

Nature Communications

Yuntian Wang, Yuhang Li, Tianyi Gan, Kun Liao, Mona Jarrahi, Aydogan Ozcan

Universal Nonlinear Optical Computing with Single-Layer Encoder-Decoder Co-Localization (Oral)

Oct 2026

Frontiers in Optics (FiO)

Universal Adversarial Microscopic Perturbations Reveal Transferable Vulnerabilities in Pathology Foundation Models (Oral)

Apr 2026

Optica Biophotonics

Diffractive Optical Processors for 3D Vibration Sensing (Oral)

May 2026

CLEO

Diffractive Waveguides Designed by Deep Learning (Oral)

Aug 2025

SPIE Optics + Photonics