Case Study: Patreon saves nearly $600k/year in ML resources with Baseten

Patreon saves nearly $600k/year in ML resources with Baseten

Patreon, the creator monetization platform serving over 250,000 creators and 8 million monthly patrons, needed a faster, more cost-effective way to transcribe large volumes of audio and video and add auto-generated captions to its product. As its ML needs grew, the team wanted to avoid spending excessive time and engineering effort on infrastructure while meeting security and compliance requirements. Baseten provided the ML infrastructure for serving the Whisper transcription model.

With Baseten, Patreon deployed and scaled Whisper quickly without building an in-house ML infra team. The result was 440+ hours of dev time saved per year, about $600k in annual resource savings, and 70% lower GPU costs. Patreon also said Baseten was roughly half the price of using OpenAI for Whisper, while improving control, security, and compliance.

Patreon

Nikhil Harithas

Senior ML Engineer