Aicadium Accelerates AI Deployment and Governance for Enterprises on AWS

The artificial intelligence (AI) industry is expected to be worth $190 billion by 2025. Advancements in AI have improved the way we live, work, and relate to each other. As a result, interest in Responsible AI—an ethical framework and set of principles that guides businesses designing and deploying AI models—has increased.

To address this, Aicadium (formerly known as BasisAI)—a Singapore-based global technology company that manages the development of real-time, scalable, and compliant AI systems from algorithms to architecture—launched Bedrock. An end-to-end machine learning (ML) platform, Bedrock helps data-driven companies govern their AI systems and deploy fair, trained models in production. It has built-in features to abstract elements—such as monitoring and auditing—away from the ML deployment process to avoid bias.

BasisAI launched Bedrock on the Amazon Web Services (AWS) Cloud in 2019 and became an AWS Partner the following year. In August 2021, it was acquired by Temasek-founded Aicadium and the two organizations became jointly known as Aicadium. “Most of our customers use AWS or want to,” explains Linus Lee, vice president of Data Science at Aicadium. “AWS products are tried and tested in the market and are very robust, which reduces technical risks such as having to constantly adapt backend APIs to new service updates.”

Choosing the cloud over an on-premises infrastructure was an easy choice for a startup with limited initial funding, Lee adds, as running Bedrock out of an on-premises data center would have doubled its costs. To run cost-effective ML workloads with high performance, Bedrock relies on Amazon Elastic Compute Cloud (Amazon EC2) GPU-based instances such as G4dn. The platform also uses a range of automated, managed services on AWS to abstract away tedious tasks for its customers such as building docker images and converting them into containers. This allows ML engineers to focus on deploying their models in production.

In addition to government organizations, Aicadium works with multinational corporations in the finance, retail, and mobility sectors. Before securing these customers, Aicadium conducts a proof-of-value phase in which its engineers customize a sandbox environment on AWS to address various use cases. For example, a typical use case in retail would be monitoring and improving the speed of returning user search queries for relevant items in a recommender system.

After the proof-of-value phase, Aicadium moves quickly to deploy Bedrock for its customers. To streamline and speed up the deployment process, Aicadium leverages AWS built-in integrations with Terraform infrastructure-as-code software and other open-source tools. The startup also relies on Amazon Elastic Kubernetes Service (Amazon EKS) to efficiently run its AI services in a containerized format. “We can deploy Bedrock for a new customer in just one to two hours, including checks. Deploying manually would have taken us up to a week,” Lee says.

In addition to reliance on managed services such as Amazon EKS, Aicadium opted for a multi-tenant environment on AWS to deliver Bedrock as a software as a service. Its engineers can manage operations from a single control plane that connects to its customers’ Amazon Virtual Private Cloud (Amazon VPC) accounts. This approach not only saves time but is also cost effective. “By using a multi-tenant approach on AWS, our onboarding process is fast, costs are controlled, and security protocols are standardized,” says Lee.

Because AWS security and compliance checks are based on industry best practices, such as those set by the Center for Internet Security (CIS), Aicadium is able to meet its customers’ expectations of reliable security benchmarks.

Before implementing Bedrock, a majority of Aicadium’s customers performed compliance checks on a quarterly basis, which is the standard practice for enterprises in most industries. Now, Aicadium’s customer applications are scanned for compliance around the clock using AWS Security Hub.

“If we implement a new feature that isn’t compliant, AWS Security Hub notifies us within 24 hours, allowing us to remediate early. AWS helps ensure that our systems are as secure and compliant as possible at all times,” Lee says.

Aicadium plans and implements security controls as early as the development phase, using AWS Identity and Access Management (IAM) to control access rights for its multi-tenant architecture. It also uses Amazon Simple Queue Service (Amazon SQS) and Amazon Simple Notification Service (Amazon SNS) to transmit data from its customers’ infrastructures back to its central control plane. This prevents compromising sensitive data during transmission. “Building our infrastructure on AWS tightly integrates security with our application rather than around it,” says Lee.

Aside from security, cost-control, and a faster deployment time, Aicadium cites high availability as a leading benefit of the AWS Cloud. The startup uses two to three Availability Zones across each AWS Region to ensure fault tolerance. It also recently implemented Amazon Auto Scaling Group (Amazon ASG) instance refreshes and Amazon Elastic Container Registry (Amazon ECR) cross-account replication to improve application reliability, which ensures an availability SLA of 99.9%.

These new service features, and the pace at which AWS introduces them, help Aicadium enhance Bedrock over time. “Services such as Amazon EKS just keep growing and improving. AWS manages the evolution of its products seamlessly and continues to introduce new features that facilitate innovation and help us manage our platform efficiently,” says Lee. With its foundations built on AWS, Aicadium is now looking at expanding into markets outside the Asia Pacific region and increasing its number of customers in the financial services industry.

To learn more, visit Amazon Security Hub.