Technology ID
TT-0008
Description
We provide efficient edge computing chips and systems that can handle large AI models in real-time, creating efficient, low-cost, and flexible intelligent inference hardware to provide the core “brain” for embodied AI, driving the development of applications for embodied AI. We utilize a software-hardware co-design platform, model compression techniques, and advanced digital compute-in-memory technology to address challenges such as real-time processing, privacy, and cost in handling large AI models and embodied intelligence computing, while meeting the demands of heterogeneous computing, energy efficiency, and storage bandwidth.
Content
Customized AI hardware design using application-algorithm-hardware co-design platform
Digital Compute-In-Memory technology for high power efficiency AI computing
Compression technology specifically designed for transformer-based large AI models
Inventors
Applicable Industries
Potential Application
Technology Focus
TRL
Images
