IT Infrastructure
At Christian Gloor Computational Science, projects are backed by a robust and modern in-house compute environment purpose-built for machine learning, data science, and AI engineering workloads.
The infrastructure consists of a 12-node high-performance compute cluster, each node equipped with multi-core CPUs (in total over 300 Cores at 5 GHz and over 1 terrabyte of RAM) and NVIDIA GPUs up to the RTX 5090. This eliminates the typical bottlenecks of cloud provisioning and shared environments, enabling rapid experimentation and fast iteration cycles for research and development.
- Fully containerized orchestration using Kubernetes, optimized for reproducible ML workflows and horizontal scaling
- Isolated dev/test/production environments to ensure separation of concerns and safe iteration cycles
- Support for GPU-accelerated libraries and frameworks including TensorFlow, PyTorch, XGBoost, RAPIDS, and CUDA-native tools
- 10 Gbit/s fiber-optic internet uplink for fast data transfer and cloud/service integration
- 10 Gbit/s internal interconnect between nodes for low-latency, high-throughput parallel computing
- Geographically separated backup server for full redundancy and disaster recovery
