Is this your business? Claim it to manage your IP and profile
EdgeThought by Skymizer focuses on bringing high-efficiency AI inferencing capabilities directly to edge devices. This IP is centered around a compiler-driven, software-hardware co-design that ensures optimal resource efficiency in executing large language model (LLM) inferences. Engineered to minimize hardware demands, EdgeThought’s architecture is compact yet powerful, making it ideal for use in constrained memory environments. EdgeThought's dynamic decompression engine is a hallmark feature, facilitating on-the-fly model weight decompression which reduces both storage requirements and memory bandwidth consumption, all while maintaining high inference accuracy. This approach enables EdgeThought to enhance execution efficiency without the need for expensive, state-of-the-art hardware, making cutting-edge AI more accessible and cost-effective. Built on the robust LISA v2 and v3 architectures, EdgeThought integrates seamlessly with existing AI ecosystems, supporting popular LLM frameworks like HuggingFace and OpenAI APIs. This integration is complemented by a broad toolkit that includes tools for finetuning and retrieval-augmented generation, underscoring EdgeThought’s adaptability in various AI applications from IoT devices to high-performance edge servers.
HyperThought represents a leap forward in AI IP design by Skymizer, purpose-built for large language models (LLMs) to maximize performance and power efficiency at the edge. Its design features advanced compression technologies that significantly reduce language model size, resulting in lower parameter counts and diminished DRAM bandwidth needs. HyperThought operates efficiently with LPDDR4/5 memory, minimizing footprint without compromising on model integrity. The architecture of HyperThought is built for balanced performance, combining high throughput capabilities with optimal area usage to achieve unmatched compute efficiency. It delivers robust performance even on a 28nm process node, pushing the boundaries of what's feasible in compact AI solutions. Its scalable architecture supports multi-core configurations for heightened processing power, while also allowing multi-chip integration to handle large-scale model requirements effectively, reaching up to 1200 tokens per second for expansive models. Security is integral to HyperThought’s design, incorporating the Language Instruction Set Architecture (LISA v3) to safeguard all operations. This makes it a foundational component for next-gen AI innovations, ensuring both versatility and protection in diverse AI applications. HyperThought thus stands out as a comprehensive platform optimized for the future of AI processing.
Join the world's most advanced semiconductor IP marketplace!
It's free, and you'll get all the tools you need to discover IP, meet vendors and manage your IP workflow!
No credit card or payment details required.
Join the world's most advanced AI-powered semiconductor IP marketplace!
It's free, and you'll get all the tools you need to advertise and discover semiconductor IP, keep up-to-date with the latest semiconductor news and more!
Plus we'll send you our free weekly report on the semiconductor industry and the latest IP launches!
To evaluate IP you need to be logged into a buyer profile. Select a profile below, or create a new buyer profile for your company.
