Case Study: ST Unitas achieves 10x AI scalability and 80% reduction in DevOps time with Intel Tiber AI Studio

Managing GPU HPC environment for Data scientists

ST Unitas, a fast‑growing EdTech company serving over 3 million students and owner of The Princeton Review, faced major scaling challenges as its AI team deployed personalized learning services like STELLA and CONECTS. Managing a hybrid HPC environment of 40+ Titan RTX GPUs and maintaining dependencies consumed more than half the team’s time, slowing research and model delivery. Intel Tiber AI Studio was engaged to streamline ML infrastructure and reduce operational overhead.

Intel Tiber AI Studio implemented a unified ML platform with automated DevOps, Kubernetes orchestration, meta‑scheduling across on‑prem and cloud resources, and one‑click launch of optimized PyTorch/TensorFlow containers and reproducible pipelines. As a result, ST Unitas scaled data science activity 10x in months without new hires, cut DevOps time by 80% per job, achieved 100% on‑prem GPU utilization with cloud bursting, and saved over $1M in wasted GPU and human costs; production QA runs under 2 hours while maintaining >80% accuracy. Intel Tiber AI Studio’s solution freed researchers to focus on models and faster innovation.

ST Unitas

Derrick Cho

AI Director