Ninestars achieves 10x scale on AWS by modernizing enterprise automation with AOTM's agentic AI

Ninestars supports complex, high-volume content and data automation workflows for global businesses. As AOTM took on more advanced automation workloads, the company determined that its next stage of growth would require a more scalable way to run the platform.

Its on-premises virtualization stack had supported the business for years. Over time, aging infrastructure, static provisioning, and manual operations made it harder to scale efficiently, maintain consistent service levels, and manage infrastructure costs. The environment worked well for more predictable workloads. But this was not designed to scale dynamically for enterprise workloads or support more advanced AI workloads efficiently. 

To prepare for continued growth, Ninestars set out to migrate and modernize its flagship platform, AOTM, on AWS. 

By modernizing with AWS, Ninestars changed how AOTM runs and scales enterprise automation workloads. The company scaled 10 times beyond its legacy environment, building a more dynamic model that responds faster to changing demand. Ninestars also improved service level agreement reliability from 92 percent to 99.7 percent, and set recovery objectives within 1 hour. At the same time, the move reduced total cost of ownership (TCO) by 60 percent, helping the company run AOTM more cost-efficiently as it prepared for continued platform development.

Ninestars took a phased approach to migration and modernization. About 80 percent of its workloads had been running on premises. The company set out to migrate and modernize 79 implementations running on 6,000 virtual machines. Ninestars used the AWS Migration Acceleration Program (MAP) 2.0, which helps customers accelerate migration and modernization on AWS, to support the transition. Alongside this, the company also utilized its own orchestration model to automate more of how AOTM runs, scales, and recovers workloads. The company began with proofs of concept and a pilot migration before moving larger workloads to production. This phased approach helped Ninestars reduce risk while reshaping AOTM for a more dynamic operating model. 

At the center of the shift, the company deploys Amazon EC2 Spot Instances, which provide spare Amazon Elastic Compute Cloud capacity at lower cost. Ninestars scales across 25 instances through Amazon EC2 Auto Scaling, which adjusts compute capacity automatically to meet different workload needs while managing costs. The company also uses its own orchestration model to automate more of how AOTM runs, scales, and recovers on AWS. This design helps Ninestars run more than 90 percent of compute on Spot Instances and handle high-volume automation workloads. 

 

The company also turned to Amazon Elastic Compute Cloud (Amazon EC2) to provide secure and resizable compute capacity in the cloud, across multiple AWS Regions and Availability Zones. This multi-region setup strengthens performance and recovery. Ninestars supports this environment through Amazon Elastic Block Store (Amazon EBS), which delivers block storage for Amazon EC2, and Amazon Simple Storage Service (Amazon S3), which offers scalable object storage. Together, these services help the company process 6 million graphics-intensive tasks each day and support large-scale content and data workloads.

As Ninestars scaled, it strengthened security and observability across the environment. The company uses Amazon GuardDuty to monitor threats, AWS Security Hub to centralize security findings, and AWS WAF to protect web applications. In addition, it deploys Amazon CloudWatch to monitor applications and infrastructure as part of the environment. These services help the company maintain performance, security, and visibility as workloads grow.

Ninestars plans to expand how it runs AOTM on AWS. The company has further implemented Amazon Elastic Kubernetes Service (Amazon EKS), a managed Kubernetes service, and Karpenter, an open-source Kubernetes cluster autoscaler. Together, these services can automate compute provisioning in containerized environments. 

Ninestars has also integrated Amazon Bedrock foundation models into AOTM to support advanced generative AI, intelligent orchestration, and hybrid inferencing alongside its domain-specific language models (DSLMs). Using Amazon Bedrock, Ninestars powers enterprise agentic AI workflows, scalable inference, and feedback-driven model improvement across regulated and high-volume automation use cases. This architecture helps the company to optimize performance, resiliency, and inference economics while continuing to strengthen its AI-native platform capabilities.

"With AWS, we've transformed AOTM into an AI-native platform that can deliver enterprise-scale automation through agentic AI, domain-specific models, and Amazon Bedrock. This foundation helps us innovate faster, scale globally, and deliver intelligent automation more efficiently for our customers," said Prashant Akhawat, Chief Technology and AI Officer (CTAO), Ninestars Information Technologies.