Case Study: Gepetto achieves faster cold starts and lower infrastructure costs with Beam

How Gepetto Achieved Faster Cold Starts While Cutting Infrastructure Costs

Gepetto, an AI-powered virtual staging and interior design app, faced significant challenges with its previous AI inference provider, Replicate. The customer experienced a sudden and dramatic increase in costs due to a new billing policy that charged for idle time, not just prediction time. Further pain points included unacceptably slow model boot times of up to five minutes, security concerns from using public models, and frequent service downtime that directly impacted their users.

By migrating to Beam, Gepetto achieved a superior solution. Beam provided faster cold boot times, cutting them by 5x compared to other providers, and eliminated charges for idle time, saving the customer thousands of dollars monthly. The platform also delivered greater than 99.9% production uptime and enhanced reliability, enabling Gepetto to provide a better, more stable service for its customers.

Gepetto

Simon Brami

Founder