AI Starter Kits is a comprehensive set of FPGA-based prototyping tools including optimized Open Neural Network Compiler (ONNC), drivers (KMD/UMD), Linux kernel, virtual platform, FPGA netlist and RTL code (NVDLA). Using AI Starter Kits, you will benefit from an optimal design experience with our extensive suite of advanced design tools, a welltested open source solution, and a fast-time-to-market prototyping FPGA board, all in one platform. AI Starter Kits optimize your design process and ensures that your software runs and exploits the full power of the underlying AI chips. Skymizer aims at providing a unique solution tailored to your needs to help you build a strong hardware/software co-design team and efficiently save your time ahead of AI chip fabrication.
Full-fledged NVDLA compiler
By diving deeply into NVDLA RTL code, we successfully make ONNC supports complex operators and multiple hardware configurations of NVDLA. It supports more ONNX models; supports more hardware configurations; and produces high performance loadable files.
Best supports of ONNX models
ONNC supports most popular models in ONNX model zoo. Weight of official ONNX models is not designed for NVDLA hardware, and the result may not be always correct. To resolve this problem, we create ONNC calibration tool to adjust weights in models.
AI-as-a-Service
Nowadays, web-service providers, such as Netflix or Amazon, tend to use micro-service as their way to develop new services and provide their services in the same way. But when they provide AI services to their customers. They either provide too high-level solution, such as Google AI Hub, or too low-level solution, such as Amazon SageMaker. The market needs a middleware solution of AI micro-services. We will provide this solution, AIaaS, to our customers. Customers can access AI computing power in many ways, such as sending pictures and asking whether it is cat or not and also can configure requests more precisely. Without understanding convoluted knowledge on how to use Deep Learning Accelerators(DLAs) and where to deploy these DLAs, customers can focus on program logic itself and put others to us.
Lead a team to deliver SOLL, a Solidity/Yul to Ewasm compiler for blockchain domain, and deliver SSVM, a high performant and flexible WebAssembly virtual machine for both AI and blockchain applications.