Tutorial
AssemblyAI cuts speech-to-text inference costs with AWS EC2 Spot
AWS published a case study detailing how AssemblyAI runs GPU speech-to-text inference on Amazon ECS using EC2 Spot Instances, an On-Demand baseline, and SQS-based retry handling to absorb interruptions. The architecture diversifies across g4dn, g5, and g6 GPU families and uses independent regional deployments for data residency. AssemblyAI reports 70% inference compute cost savings versus On-Demand pricing while maintaining SLA commitments as compute demand grew nearly 4x year-over-year.
AssemblyAIAWSSpeech-to-text inference