Streamline Your AI and Data Science Environments with Spack

As AI and data science workflows evolve, so does the complexity of their software environments. From CUDA and C++ libraries to Python and R packages, developers must balance performance, reproducibility, and portability across different systems and hardware.
Spack, an open source package manager developed for high-performance computing, provides the flexibility and control to manage these challenges. It enables developers to create reproducible, customizable, and shareable environments that integrate seamlessly across platforms, compilers, and GPUs.
In this webinar, Kitware’s experts will demonstrate how to design and deploy optimized AI and data science environments using Spack, including how to:
- Configure and build environments on Linux and macOS that take advantage of system CUDA drivers.
- Use Spack’s dependency resolution and build caching to accelerate setup and deployment.
- Integrate AI frameworks like PyTorch, TensorFlow, and cuDF for efficient machine learning development.
- Run a complete, reproducible workflow from training to containerized inference using the same environment definition.
By the end of the session, attendees will understand how Spack can streamline environment management for large-scale research, simplify collaboration, and ensure reproducibility across teams, systems, and projects.
Attendees will learn how to:
- Understand the value of reproducible AI environments.
- Simplify complex dependency management.
- Accelerate setup and deployment.
- Customize environments for your hardware and frameworks.

from Spack experts on how to transform your data science workflows!