Success Story
AWS MAP assessment and AI optimization for Grego AI: TCO, token-quota management and per-model cost observability
Grego AI is a generative-AI startup (with operations in Argentina) that runs language models on Amazon Bedrock and OpenAI. Craftech supports it with an AWS-funded assessment (Migration Acceleration Program) and optimization work on its AI consumption on AWS: TCO, token-quota management and per-model cost observability.
Challenge
Grego AI needed to organize and optimize its generative-AI operation on AWS: understand its total cost of ownership (TCO), manage token-quota limits with AWS, gain visibility into per-model consumption cost, and prepare an AWS-funded migration within tight deadlines.
Solution
Craftech structured an AWS MAP assessment: discovery documentation (current models and usage review, token-consumption analysis, critical-prompt identification and evaluation-metric definition), a total-cost (TCO) report and a migration roadmap. In parallel it carried out technical work: configuring an EC2 instance for a test environment, managing token quotas with AWS, defining new quota limits, and building a CloudWatch cost dashboard by model consumption.
Results
Grego AI has a full diagnosis of its AI operation (TCO, token consumption, critical prompts and evaluation metrics), a migration roadmap and tighter cost control thanks to the per-model consumption dashboard in CloudWatch and token-quota management with AWS.
Stack & technologies
