Lego Inspired accessible and automated design framework for demanding Edge AI Systems

August 2024 – May 2026

Objective
 A Lego-inspired design framework called SiLago (Silicon Lego) enables automation from the system level to ready-to-manufacture solutions for high-performance Edge AI applications. This framework bridges the gap between ease of use and performance by providing ASIC-comparable efficiency while achieving significantly improved energy efficiency—10X to 100X better than commercial off-the-shelf (COTS) solutions such as GPUs and FPGAs. The research project aims to enhance the SiLago framework to support comprehensive system-level implementation by addressing computation, storage, and interconnect requirements. These enhancements will enable SiLago to streamline the synthesis of complex applications, such as those required in industrial use cases. Finally, the improved system-level capabilities will be seamlessly integrated with the existing application-level synthesis flow, creating a unified, automated design process from applications to manufacturable silicon.

Background
The field of electronics and VLSI has driven transformative advancements in computing, enabling the development of increasingly powerful and efficient hardware systems. System architecture plays a crucial role in defining the structure and interaction of hardware components, ensuring efficient computation, storage, and communication. Despite these advancements, designing high-performance and energy-efficient hardware, such as ASICs, remains a complex, resource-intensive process requiring specialized expertise. The SiLago framework builds on these foundations, combining principles of VLSI, hardware modeling, system architecture, and design automation to provide a modular, automated solution for ASIC design.

About the Digital Futures Postdoc Fellow
Nooshin Nosrati completed her doctoral research in Digital Electronic Systems at the University of Tehran (UT). Her doctoral thesis was on hybrid reliability provisions in embedded systems with a focus on Computational Elements. Her research interests encompass hardware design and modeling, computer architectures, reliability and testability of embedded systems.

Main supervisor
Ahmed Hemani, Full Professor, Department of Electrical Engineering, KTH.

Co-supervisor
Artur Podobas, Associate Professor, Division of SCS, School of EECS, KTH.