Cloud-Native Planetary Scale Supercomputing
Spatial-temporal data is intrinsically large-scale. The amounts of both meteorology and remote sensing data are at PB level and are still growing at TB-level on a daily basis.
TerraQuanta's computing infrastructure enables AI to efficiently deliver accurate forecasting results and surface model anywhere in the globe.
PB-Level Data Processing
Processing up to 100 billion data points and inhaling TBs of data on a daily basis, effectively support large-scale and high-frequency computing mission.
Cutting-edge MLOps
Neural Architecture Search (NAS), AutoML, and CUDA-based acceleration are applied in distributed training and processing.
Grow With Giants
Proudly graduated accelerators by Microsoft, Intel, and Nvidia, TerraQuanta has abundant cloud and in-house computing resource.
OceanStorm
Management Platform For Smart Large-scale Computing
A computing management platform for spatial-temporal data.
Flexibly scheduling thousands of GPUs and CPUs
Execute billions of data point calculations, hugely supporting AI training and processing of globe-scale weather forecasting
Poseidon
GPU Management Engine for Model Training and Inference
Enables distributed training and large-scale data manufacturing, constantly integrates and deploys deep learning models in their complete life cycles.
In combination with GPU technologies such as TensorRT, dramatically accelerates data processing
Beyond deep learning - Various GPU-based applications such as FFT
Halia
CPU and Model Management Engine For Parallel Computing
Digital earth is more than deep learning. Analytical and numerical models also rely on CPU cluster
Use high performance CPU cluster to regularly execute large-scale numerical computing
Cloud-native architecture for constant integration and deployment of analytical and numerical models
