Wangsu Technology launches one-stop edge intelligence solution to lower AIGC application threshold

The new generation of intelligent technologies represented by AIGC (generative artificial intelligence ) and big models have triggered a new wave of innovation around the world . With the emergence of big models , how to make AIGC more convenient and widely applied to vertical scenarios has become a focus of the industry.

Recently, at the Global Edge Computing Conference, Wangsu Technology announced that it has launched a one-stop edge intelligence solution for AIGC . From the three levels of Serverless GPU, AI gateway, model fine-tuning and inference services , it greatly reduces the implementation threshold of AIGC applications, aiming to help users easily realize AI application innovation in a low-cost and efficient manner .

Chen Yunhui, senior architect of Wangsu Technology's edge intelligence platform

Chen Yunhui, senior architect of Wangsu Technology's edge intelligence platform, pointed out at the meeting that although there are many basic large models in China, providing users with a rich selection of models, users still face high barriers in actual implementation.

Specifically, in the process of implementing AI applications from scratch , enterprise users must comprehensively consider multiple key factors , including comparing the inference cost and performance evaluation of open source models and cloud service providers, paying attention to system security, sensitive word filtering , etc., especially in the transformation and integration with existing businesses, there are many problems . For application scenarios with strict requirements on data privacy and model control, it is also necessary to consider fine-tuning private models for specific vertical scenarios. All of these make the entire implementation cycle long and costly .

Edge computing has advantages such as efficient data processing capabilities , real-time response speed, and lower bandwidth requirements, and can play a key role in the application of AIGC . The general consensus in the industry is that the reasoning and content production of AI models in the future will be distributed in a gradient manner on the cloud side, edge side, and terminal side.

Relying on its deep accumulation in the field of edge computing, Wangsu Technology has continued to explore AIGC in recent years, and has gradually formed a one-stop edge intelligence solution covering Serverless GPU, AI gateway, model fine-tuning and inference services .

According to reports, Wangsu Serverless GPU provides users with highly scalable, on-demand GPU computing resource services, which are mainly suitable for a variety of computing-intensive tasks, such as AI model reasoning, audio and video acceleration production, graphics acceleration, etc. Based on Wangsu's rich GPU computing resources, combined with the advantages of Serverless K8S and GPU acceleration, users do not need to manage GPU servers and K8S, but only need to upload container images to run AI applications conveniently and quickly ."In addition to the feature of free operation and maintenance,Serverless GPU has the advantage of low latency compared to centralized deployment becauseGPUresources are placed on the edge, closer to users and closer to the business edge. In addition, through on-demand allocation andGPU virtualization technology,the utilization rate of GPU resources isgreatly improved,which caneffectively reduceusers' GPU usage costs." said Chen Yunhui.

Wangsu AI Gateway is a unified API interface for large language models. It can be used between applications and hosted large language models (LLMs), allowing API requests to be forwarded to various cloud vendors and self- deployed large models through a unified API interface . Wangsu AI Gateway can be used for individual developers and enterprise-level AI applications to help manage and integrate multiple backend large model interfaces, improve the stability and reliability of AI applications , and reduce the cost of calling backend models .

In addition, companies that have strong requirements for data privacy and model controllability often choose to deploy their own open source models. In this regard, Wangsu's edge model training and inference platform can provide the most economical and reliable model inference service.

Wangsu's edge model training and inference platform not only provides users with a global edge inference operating environment, but also provides a model fine-tuning environment, equipped with the required GPU resources and a complete tool chain, and supports one-click deployment of fine-tuned models to Wangsu's global edge nodes. Through continuously optimized inference performance, the edge model training and inference platform can support users to efficiently train, infer, and use large models.

At the meeting, Chen Yunhui also talked about the development of AIGC . He believed that big models will evolve into infrastructure in the future, just like the current operating system and big data platform. Although users may not be aware of it, it provides basic and universal support for various applications. At the same time, open source capabilities will become stronger and stronger.

It is worth mentioning that in addition to the above three products, Wangsu Technology is also actively exploring two new directions: AIGC workflow engine and AI marketing creative platform to further enrich its edge intelligence solution capabilities .

Among them, the AIGC workflow engine is designed to help users manage and execute the workflow of AI applications more efficiently . By breaking down the current complex business processes into individual small tasks and reducing the complexity of a single task, it can not only reduce the reliance on prompt words and large model reasoning capabilities, but also improve the performance and fault tolerance of large models in handling complex tasks. At the same time, in the links where the performance of large models is insufficient, the AI ​​workflow engine can also use traditional processing methods. The AI ​​application development platform built on the workflow engine can effectively combine AIGC with traditional methods to better implement the implementation of AI applications. The AI ​​marketing creative platform provides end users with out-of-the-box AI products , aiming to empower users' marketing work.

"In the future, we hope to continue to provide customers with edge intelligence solutions with lower-cost model inference and higher-quality content generation," said Chen Yunhui.