Building Efficient ML Systems and Multimodal Intelligence.

Ph.D. Student in Computer Science
Purdue University

I am interested in algorithm-hardware co-design for efficient generative models, multimodal models, and world models.

Email GitHubGoogle ScholarLinkedIn

Jianwen Wu

01 / 05

Education

2026 — Present

Purdue University

Ph.D. Student in Computer Science

Research focus: efficient and multimodal intelligence.

West Lafayette, Indiana

2022 — 2026

Nanjing University

B.Eng. in Integrated Circuits and Integrated System

Nanjing, China

Suzhou, China (Suzhou campus)

02 / 05

Experience

2026 — Present

West Lafayette, Indiana

Graduate Research Assistant

EcoAI Lab · Purdue University

Exploring efficient learning and generation across models, modalities, and computing systems.

Efficient AI Systems
Multimodal AI
Generative Models

2025 — 2026

South Bend, Indiana

Visiting Undergraduate Research Intern

Hardware-Software Codesign Lab · University of Notre Dame

Research across efficient learning algorithms, specialized hardware, and neuro-symbolic systems.

Efficient AI Systems
Algorithm-Hardware Codesign
Neuro-Symbolic Inference

03 / 05

Projects

Preview of Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling

Research Paper · arXiv

2026

Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling

Lixuan Wei, Wei Zhou, Jianwen Wu, Yipeng Shen, Meiling Wang, and Haoran You

A plug-in decoding framework for diffusion LLMs that learns variable-length blocks and position-aware parallel sampling, achieving up to 18.76× higher throughput than vanilla decoding with comparable accuracy.

Diffusion LLMs
Efficient Inference
Parallel Decoding

04 / 05

Moments

05 / 05

News

SEP 2026

Joined Purdue University as a Computer Science Ph.D. student.

Beginning a new chapter in West Lafayette.

AUG 2026

Started research at the EcoAI Lab.

JUN 2026