UMNAI worked with The Scale Factory to bring their innovative AI solution to market on AWS.
Hybrid Intelligence AI
UMNAI are building Hybrid Intelligence, the world’s first neuro-symbolic AI platform. Their third-generation AI framework delivers inference results that are traceable and explainable, solving problems of transparency that are inherent to previous AI approaches.
Explainable models have huge benefits in terms of audit and compliance, making them suitable for adoption by industries that’ve traditionally been hesitant to deploy AI in consequential use cases because of concerns about risk, bias and the inability to anticipate unintended consequences with confidence.
Hybrid Intelligence models are easier to use than other types of AI tools, dramatically reducing the time to take an AI use case from conception to productive deployment, resulting in lower costs and more certain ROI. Transparent models are easier to test for correctness, which means less trial and error during the training process.
Hybrid Intelligence can cut AI Time to Production by more than 90%.
The challenge: build a production-ready SaaS Platform with Enterprise Security
With the new theory behind the Hybrid Intelligence framework tested, validated and prototyped in the lab, it was time for UMNAI to platform the product to take to the market. They recognised that, to drive adoption, it would be necessary to create a SaaS offering designed to be familiar to consumers of other AI tools.
As an AI platform handling sensitive customer data and training proprietary models, UMNAI required enterprise-grade identity and access management from the foundation. Their multi-account AWS environment needed centralised IAM governance, federated authentication for their team, and fine-grained access controls for AI workloads. With services spanning Amazon SageMaker, Amazon S3, and container orchestration, implementing least-privilege IAM policies would be critical to maintaining security whilst enabling their AI platform to function.
UMNAI’s team are specialists in the AI space, but looked to The Scale Factory for our expertise in building production-grade SaaS platforms, following an introduction by a team at AWS.
“We came to this project with an incredible knowledge deficit around building on the cloud,” said Ken Cassar, co-founder and CEO of UMNAI. “On speaking with The Scale Factory it was clear that they were extraordinarily competent in this area - we were so confident in this that we didn’t feel the need to talk with other AWS consulting partners before making the decision to work with them”.
The solution
The Scale Factory and UMNAI teams jointly designed a solution through a series of discovery workshops, each group bringing their specific expertise.
As part of the design, we identified that two of our existing solutions would help move the UMNAI project forward quickly. We implemented our AWS Control Tower for SaaS solution to provide a high standard of cloud security and governance from the outset, and we used our AWS Fargate-based container platform because this workload would use containers.
Centralised Identity and Access Management
AWS Control Tower with IAM Identity Centre - We deployed AWS Control Tower to establish centralised identity and access management across UMNAI’s multi-account environment. AWS IAM Identity Centre provides federated authentication for the UMNAI team, eliminating local IAM user accounts and implementing multi-factor authentication enforcement. Standardised permission sets define role-based access for developers, operations teams, and administrators, with temporary credentials replacing long-lived access keys.
Service Control Policies for Organisational Governance - We implemented Service Control Policies (SCPs) across the AWS Organization to enforce security guardrails and prevent privilege escalation. These policies restrict actions across all accounts, ensuring consistent security posture regardless of individual account configurations. SCPs prevent unauthorized modifications to IAM Identity Centre, enforce encryption requirements, and block risky service configurations.
Audit Logging - All IAM access is logged through AWS CloudTrail integration, providing complete audit trails for compliance and security investigations.
Platform Architecture
Architecturally, the UMNAI product consists of a Control Plane – which runs the admin APIs and helps with tenant provisioning and cost attribution – and a Workload Plane where the customer-facing solution itself runs.
Customers provide data into Amazon S3 (either using their own bucket or one provided by the platform). This data is onboarded, prepared, and used to train a Hybrid Intelligence model using UMNAI’s proprietary training method running on Amazon SageMaker.
These trained models are then made available using Inferencing Endpoints from Amazon SageMaker, optionally using Serverless Inferencing to keep costs down where appropriate.
The platform currently uses a bridge tenancy model, with some resources being shared across customers and others – most notably the data storage solution – being unique to each tenant. A silo tenancy model will be optionally available in future, for customers who are willing to pay a little more for a higher level of isolation.
Application User Authentication - Customer authentication and authorisation for the UMNAI application interface is handled by a user management platform provided by AWS Technology Partner Frontegg, which manages end-user login and application-level permissions separately from the AWS infrastructure IAM controls.
What UMNAI said...
On speaking with The Scale Factory it was clear that they were extraordinarily competent in this area
We were so confident in this that we didn’t feel the need to talk with other AWS consulting partners before making the decision to work with them
AWS: the most mature cloud platform
Before choosing AWS as their target cloud provider, UMNAI spent some time looking at the alternatives.
“We selected AWS to build on because they’re the most mature cloud platform”, says Cassar, “they clearly know what they’re doing”.
As with many of our projects with SaaS customers, this work was eligible for co-investment by their chosen cloud provider.
“The Scale Factory were very proactive and effective in helping us secure funding from AWS for our project,” said Ken.
The results
The UMNAI platform is now in production for early adopters and we’re looking forward to seeing everything scale as demand for neuro-symbolic AI increases.
