Case Study: Information Sciences Institute speeds up machine-learning research and reduces costs with Altair Knowledge Works' Grid Engine

ISI's VISTA Lab Speeds Up Machine Learning Research and Lowers Expenses

The Information Sciences Institute (ISI) at USC Viterbi is a leading research center in computing and AI whose VISTA lab runs extensive machine learning projects in areas like facial recognition and NLP. VISTA’s shared compute cluster was managed ad hoc, leaving GPUs idle and forcing researchers to spend significant time troubleshooting an open-source scheduler that couldn’t handle expensive GPU workloads or growing multi-user demands.

VISTA deployed Altair Grid Engine for its built-in GPU support, documentation, and support, enabling a project-based share tree and scalable management of 100+ NVIDIA GPUs. The platform handled parallel processing and job dependencies for terabyte-scale experiments (one processed over 3 million images), boosting research throughput, running the infrastructure at about 95% capacity with lower costs, and reliably accelerating ISI’s facial-recognition and other ML work.

Information Sciences Institute

Stephen Rawls

Programmer and Research Analyst