Case Study: Teldar achieves on-demand AI workload scaling with Zazmic

Teldar cuts GPU costs with Zazmic through scale-to-zero AI infrastructure

Teldar, a Swiss AI consultancy, faced significant financial and operational challenges managing their demanding machine learning workloads on Google Cloud. The high cost of idle premium GPU instances and the manual effort required to provision them were slowing down projects and diverting developer focus from building models. They engaged Zazmic to implement a solution that would make their high-end infrastructure an on-demand utility.

Zazmic designed and implemented an automated, event-driven scale-to-zero architecture on Google Cloud. The solution used Google Cloud Functions to automatically provision and terminate GPU-enabled VMs based on workload demand, leveraging Spot VMs for significant cost savings. This approach, defined by Terraform, provided Teldar with production-ready stability and operational autonomy. The results were transformative, delivering a significant reduction in GPU compute costs, a faster time to execution, and the ability for Teldar's team to manage the infrastructure independently.

Teldar