RP4 - Cross-layer Optimization and Demonstration for Targeted Applications
Leveraging the technologies developed in the above 3 research projects to perform cross-layer optimization and develop hardware-accelerated working systems for domain-specific, large-scale AI applications with revolutionary performance.
RP4-1:
Efficient AI Computing for Multi-modal and Large Language Model targeting at Real-Time Applications
This project will focus on applications requiring multi-modality understanding with LLM, such as healthcare and medical diagnosis, autonomous vehicles, embodied robotics, education, and learning. It will address the issue of designing highly efficient computation systems to support the computationally intensive multi-modal AI agents, through co-design of algorithms and hardware.
RP4-2:
Co-located Secure and Energy-Efficient Compute-in-Memory Hardware for Biomedical Devices and Application
This project addresses the design of next-generation smart and secure biomedical devices which will play a crucial role in diagnosis, treatment, rehabilitation, and daily health monitoring. It will further develop computing-in-memory technology to facilitate secure AI computation targeting at bio-medical applications.
RP4-3:
Scalable Architecture and System for Convex Optimization Solvers in Finance, Robotics, and Digital Economy
This project aims to design a scalable architecture and system for convex optimization solvers that target applications requiring complex convex optimization, such as financing computing, and digital economy.
RP4-4:
Domain-Specific SoC Framework for Enhanced Rendering
Performance in Immersive
Technologies This Project focuses on designing a domain-specific SoC framework to enhance 3D-rendering performance for immersive applications that will tackle the challenges of designing intelligent hardware to support real-time 3D rendering for immersive AR/VR applications.