Page 297 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

2�2 Benefits of use case

The large model toolchain platform developed by China Telecom provides full-process
intelligent services that significantly enhance the quality, efficiency, and sustainability of
communication infrastructure. 4.3 - 5G

By targeting the needs of the telecommunications industry, the platform drives innovation
through the integration of AI and large model technologies. It collects and utilizes real-time
infrastructure data to build intelligent agents that operate across knowledge management,
model training, inference, and deployment modules. These agents are tailored to meet specific
business requirements, enabling dynamic, data-driven decision-making and operational
optimization.

The platform offers advanced tools to process complex telecom data, train domain-specific
models, and deploy intelligent agents. It supports a wide range of agent construction methods
and allows flexible permission controls for collaborative use across teams or departments. This
contributes to more intelligent and efficient cloud network operations, improving resource
utilization, reducing operational costs, and addressing challenges in complex telecom scenarios.

By strengthening AI-driven infrastructure management, promoting automation, and fostering
industry-level innovation, the platform lays a strong foundation for long-term technological
advancement and more resilient, adaptive telecom systems.

2�3 Future Work

In 2025, we will strive to improve the coverage of the toolchain platform in the entire process of
model and intelligent agent construction, enhance stability, and create a toolchain platform that
can support concurrent use by 30% of users across the entire network, supporting the research
and development of intelligent agent applications in various provinces. In the future, we will
continue to build a toolchain platform ecosystem, promote the integration and collaboration
of intelligent agents in different scenarios, achieve cross industry customization, empower
thousands of industries, and build applications for the entire industry.

Next, based on this achievement, China Telecom will actively introduce new partners, gradually
expand the scale of the alliance, jointly promote the formulation of standards, application
for awards, and participation in industry evaluations, and jointly promote the sustainable
development and innovation of the network large model industry.

3 Use Case Requirements

• REQ-01: It is critical that the platform supports large-scale model training and inference
through distributed GPU clusters with scheduling, fault tolerance, and mixed-precision
optimization.
• REQ-02: It is expected that the system enables ingestion and synchronization of
multimodal data sources—including structured logs, unstructured text, images, and time-
series data—for both real-time and batch processing.
• REQ-03: It is expected that the knowledge management module allows users to upload,
manage, and preprocess datasets with support for data source integration, permission
control, and dataset versioning.

261