-->
Bridging Theory and Applications in Modern Science
Co-Hosted by DAC 2025
Building on the success of its inaugural edition, DCgAA 2025 delves deeper into the intersection of deep learning (DL) and hardware co-design, with a strong emphasis on real-world deployment and next-generation innovations in generative AI. This year, the workshop expands its focus to include key advancements such as multimodal fusion, real-time adaptive processing, and decentralized edge intelligence.
Recognizing the rise of foundation models, diffusion models, and large-scale generative systems, DCgAA 2025 prioritizes optimization for scalable, sustainable deployment. Topics will explore performance-energy tradeoffs and accessibility across diverse platforms, including edge devices, AR/VR systems, and IoT infrastructure—ensuring generative AI remains both powerful and practical.
The workshop will feature a rich mix of keynotes, paper sessions, interactive discussions, and new program elements. By uniting researchers, engineers, and innovators from academia and industry, DCgAA 2025 aims to push the frontier of hardware-aware generative AI and set new standards for scalable, efficient AI integration in the real world.
Explore cutting-edge research at the intersection of deep learning and hardware acceleration to unlock new levels of efficiency and scalability.
Engage in discussions on deploying generative AI in diverse environments—from edge devices to immersive AR/VR platforms.
Connect with leading minds from academia and industry to shape the future of hardware-aware generative AI together.
The workshop will cover a broad range of topics to illuminate cutting-edge advancements and facilitate collaboration across domains, including:
Exploring network optimization under hardware constraints for scalable generative models.
specialized hardware to meet the unique demands of generative models.
Harnessing generative models to inspire innovative hardware architectures.
Optimizing generative models for greater efficiency and application diversity.
Investigating principles and theories underpinning effective collaboration between DL and hardware.
Revolutionizing edge computing and immersive experiences in augmented and virtual reality.
Enhancing real-time and practical deployment through advanced acceleration techniques.
Learn from world-renowned experts who are at the forefront of scientific innovation
Note: The speakers are in the order of their last name.
Professor and Associate Chair of the Computer Engineering Division
University of Southern California, Viterbi School of Engineering
Topic: Efficient and Robust Personalization and Adaptation of Large Models
BioJohn Cocke Distinguished Professor
ECE, Duke University
Topic: Towards Efficient Generative AI - Software/Hardware Co-Design for the Next Generation of Intelligent Systems
BioVice-Provost (Academic Affairs) and Provost's Chair Professor of Computer Science
National University of Singapore
Topic: Adaptive Spatial Accelerators for Sparse LLMs
BioSilicon Laboratories Endowed Chair / Professor
ECE, The University of Texas at Austin
Topic: Closing the Generative AI–Hardware Loop: Photonic Acceleration, Memory-Efficient Training, and AI-Driven IC Design
BioProfessor at Computer Science Department
University of California, Los Angeles
Topic: Energy-Efficient LLM Inference
BioAdditional speakers will be announced soon. Check back for updates.
DCgAA 2025 introduces an enhanced program format featuring invited paper talks from DAC, interactive discussions, and panels. A key innovation for this year is the emphasis on dynamic and collaborative elements, such as interactive poster sessions and informal brainstorming opportunities during the coffee break.
David Z. Pan
Yizhou Sun
Yiran Chen
Peter A. Beerel
Tulika Mitra
Each talk will be 35 minutes long, followed by a 5-minute Q&A session. The schedule is subject to change, and updates will be provided closer to the event date.