Overview
Government agencies are increasingly using Machine Learning (ML) to address a broad range of mission-based needs. ML is a subset of Artificial Intelligence (AI) that uses algorithms trained on data sets to create self-learning models that can predict outcomes and classify information.
Precise’s Machine Learning as a Service (MLaaS) is an enterprise-level, cloud-based service that consists of machine learning tools and solutions.
Challenge
As demand grew for Artificial Intelligence (AI), GenAI and automation to support mission-critical initiatives, a major federal agency faced challenges in establishing centralized processes and infrastructure for securely delivering AI services at an enterprise level. Key challenges included the absence of standardized platforms for AI service delivery, lack of clear governance for infrastructure security and compliance, and the risk of fragmented or duplicative efforts across multiple business units.
Solution
Our team won an agency award for creating an enterprise MLaaS platform that integrates Cloud Service Provider (CSP) AI offerings into a unified, secure environment. Supporting the CTO office, we managed the evaluation, security assessment, design, development, and delivery of the agency’s enterprise MLaaS platform, enabling business units to securely adopt AI and automation solutions through pre-approved infrastructure, shared services, and enterprise licensing ensuring consistent governance, reducing duplication, and promoting cost-effective reuse.
To understand the agency needs and issues, our team conducted interviews with the agency’s IT teams and business units, identifying issues in the areas of people, processes, and technology. We then developed a set of services that offer ready-made, slightly generic ML learning tools that can be adapted by any agency business center as needed. These services include data visualization from forms, image recognition (e.g., recognizing pills from pictures taken on iOS and Android phones, an ML model for remediating documents to comply with Section 508 of the Rehabilitation Act to make documents accessible to those with disabilities, and numerous related Application Programming Interfaces (APIs) for natural language processing, predictive analytics, and deep learning, among others.
We developed MLaaS using REST APIs, an AWS Fargate-based serverless architecture, an OAuth 2.0 API gateway, and audit logging to validate AI model provider compliance and support human-centered AI adoption across the agency’s business units. The infrastructure and data security guardrails we implemented resulted in securing a Federal Information Security Modernization Act (FISMA) High Authority to Operate (ATO) certification for compliance purposes.
Each service is a self-contained microservice which can have regular service-to-service interaction through Amazon Web Service’s (AWS) Elastic Container Service. Our team cataloged all proposed APIs, documented the APIs, and stood up the API Hub. All requests from outside are directed through API Gateways and a load balancer that decides where services are to be directed.
Impact
Precise MLaaS empowers a federal agency to tap into ML tools and models as needed to streamline and automate their mission activities. Key benefits include:
- Establishes enterprise “as a Service” best practices via an MLaaS Knowledge Portal that provides “as a Service” best practices, AI/ML standards, and training for the agency’s user community.
- Enables collaboration and re-use of existing solutions by breaking down silos and creating efficiencies across the organization.
- Consolidates license, infrastructure and operational costs leading to overall cost savings for the agency.
- Consumption-based financial model with the the enterprise-level IT office providing ML operations and maintenance services and invoicing business units based on their utilization of available ML tools and services.
