Education & Research AI Infra

Functional Overview

Provide a closed-loop service system covering "compute power - platform - data - services - operations," supporting institute-level, campus-level, and regional intelligent computing center construction. It includes heterogeneous intelligent computing scheduling, domestic AI computing centers, large model and AI4S computing framework migration and optimization, professional domain research environments, and automated laboratories.

Customer Pain Points

  • Research computing resources are scattered, underutilized, and cannot be scheduled in a unified manner.
  • Domestic computing power (such as Huawei Ascend) is difficult to adapt, and model migration and optimization lack experience.
  • There is a lack of research environments for specific disciplines such as drug R&D, new materials, meteorology, and nuclear energy.
  • Traditional experiments are inefficient, have low automation, and require time-consuming and labor-intensive data collection and processing.

Solution Advantages

  • Heterogeneous Intelligent Computing Unified Scheduling: supports super-intelligence fusion scheduling, hybrid cloud scheduling, and multi-campus scheduling within a single university, building a scheduling-optimization-security framework.
  • Domestic AI Computing Center: provides Ascend all-in-one machines, CloudMatrix384 clusters, and public cloud computing platforms, adapted to ARM64/RISC-V/X86 and NPU/GPU/TPU.
  • Large Model and AI4S Optimization: inference optimization and training optimization, supporting model migration, performance tuning, and operator development.
  • Professional Domain Research Environment: covers disease diagnosis, drug R&D, chemical new materials, aerospace, meteorology, nuclear energy, and other fields, providing scientific research training, discipline-specific large model training, and data services.
  • Automated Laboratory: high-precision measurement equipment, inspection and testing equipment, experimental robots, and lights-out laboratories, improving experimental efficiency and data quality.

Architecture Diagram

Architecture Diagram