What challenges did Cresta face before AI implementation?

Multi-cloud ML pipelines were tedious to manage and increased security and compliance effort. Moving datasets and trained models between cloud environments slowed releases and made it difficult to keep data current. Cloud boundaries caused workflow breaks and made resource optimization and scaling harder.

Source

Evidence: Low25/100

Cresta migrates customer support conversational ML workloads to AWS (EKS/EC2/ECS, Aurora, S3)

Use case typePredictive decision supportUpdated Jun 13, 2026

Cresta is a California-based AI startup that builds real-time coaching and management solutions for sales and service teams. The company consolidated its machine learning workloads on AWS after previously running multi-cloud pipelines for dataset storage, model training, validation, and production inference. The post describes Cresta's production NLP models for suggestions, smart compose, intent classification, and named entity recognition, along with the AWS-based training and serving architecture.

Organization: Cresta
Industry: Professional Services
Location: United States
Published: December 2021

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Cresta
Provider: AWS
Maturity: Unknown
Linked source: AWS Machine Learning Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 3

1Machine Learning Operations
2Predictive Analytics
3Customer Service

Cresta moved training and inference workloads to AWS and consolidated dataset generation, training, validation, and serving in one environment.
Amazon Aurora PostgreSQL stores training data, Amazon S3 is used for snapshots and model artifacts, and Argo Workflows runs on Amazon EKS to orchestrate ML jobs.
Training uses Amazon EC2 Spot Instances and GPU-backed EC2 instances, while production inference uses PyTorch TorchServe on Amazon EKS across multiple Availability Zones.

Technologies

Amazon Aurora PostgreSQL Amazon S3 Amazon EKS Amazon EC2 Spot Instances Amazon EC2 PyTorch TorchServe Argo Workflows

Teams no longer manage ML pipelines across separate clouds, improving productivity.

Architecture

Cresta migrated from a multi-cloud ML setup to AWS. The new architecture stores datasets in Amazon Aurora PostgreSQL, stages and stores artifacts in Amazon S3, orchestrates parallel training jobs with Argo Workflows on Amazon EKS, trains with GPU-backed Amazon EC2 instances and Amazon EC2 Spot Instances, and serves production models using PyTorch TorchServe on Amazon EKS across multiple Availability Zones. Dataset generation, training, validation, and inference all run in the same AWS environment.

Sources & evidence1

Evidence: Low25/100Evidence strength