AISWare AI² is an infrastructure used to build large-scale intelligent services. It is the hub of the middle-ground of companies, which can provide step-by-step construction of algorithm models and full life cycle management services for the intelligent transformation of companies. It can help companies continue to deepen their own businesses to obtain algorithm models and achieve the purpose of reuse, combination of innovations, and large-scale construction of intelligent services. It can provide strong scenarios, low-cost, integrated software and hardware solutions and standardized AI applications for the intelligent transformation of companies.
Four specifications of engine versions of Nano, Mini, Standard, and Jumbo are provided according to the deployment requirements of different intelligence integration scenarios.
Being compatible with a variety of data annotation tools; supporting the access to big data platforms, and meeting the data reading requirements of large data training and inference tasks by means of sharding reading, distributed processing, etc.
Supporting three model training modes: wizard, drag-and-drop, and encoding. Supporting the development of distributed model training tasks and being compatible with multiple computing frameworks.
Providing a standard and universal model management interface to manage the AI models developed internally and externally in a compatible manner.
Supporting the construction and automated release of online and offline inference services. Supporting TensorRT and OpenVINO inference accelerator engines.
Providing the capability to control the resources, calls and traffic of AI services to ensure the safe, stable and efficient operation of AI middle-ground.
Diversified modeling methods.Providing diversified custom modeling methods (wizard, drag-and-drop, and encoding) to meet the modeling needs of people with different abilities.
Flexible deployment method.Providing multiple model output methods such as SDK and API, which can be quickly and seamlessly integrated with existing platforms and systems of customers.
Able to be integrated with third-party AI capabilities.Integrating third-party AI capabilities with the AI
Many achievements have become international standards.The AI
Presetting industry excellent models and related characteristic indexes to form industry-wide index sets and index processing method, and improve modeling efficiency.
Intelligence aided modeling.Freeing users from paying attention to the algorithms and parameters used, the platform can automatically select according to the data, which is more scientific and efficient.
Lowering the barriers for developing AI services
Reducing the cost of using AI to integrate intelligence
Improving the quality of building AI services
General AI services and industry models
Due to the problems such as inconsistent data standards, inconsistent service scheduling, inconsistent authority management, failure to link up end-to-end processes, failure to share platform resources, and lack of data protection standards resulted from the past “shaft-style” AI capability building, the provincial operator was in urgent need of constructing a set of one-stop end-to-end digital intelligent middle-ground.On the basis of AsiaInfo’s AISWare AI², an AI data middle-ground has been built to provide full-process AI development capabilities, integrate data middle-ground and technical middle-ground, realize the global empowerment of B/O/M, and provide multiple intelligence integration methods and multiple deployment methods, thus realizing the comprehensive empowerment of different scenarios, and the empowerment for the full process from perception to cognition to decision-making.
In recent years, the group has attached importance to the centralized construction of machine learning capabilities, aiming to reduce the overall construction cost of machine learning capabilities for the company through unified construction. Based on this guidance, the group has built a set of unified centralized machine learning platform with the help of AsiaInfo’s AISWare AI², enabling various provinces or professional companies to enter and use the platform in the form of tenants. Through effective multi-tenant management, model sharing viewing, model sharing use, model sharing recycling are realized to reduce the closure level of model and increase the penetration rate of excellent models.