GovTech Collaborates with AWS Generative AI Innovation Center to Scale Generative AI Adoption Across Public Sector Organizations

GovTech identified generative AI as a way to drive innovation, enhance public services, and improve lives within the community. However, the high costs of operating LLMs at a government scale were key considerations for the organization.

To address cost concerns and tap into the potential of generative AI, GovTech collaborated with the AWS Generative AI Innovation Center to develop generative AI solutions that enable cost-effective adoption of this transformative technology across government agencies.

To build and deploy MAESTRO, GovTech worked with AWS, including the AWS Generative AI Innovation Center, a program that pairs organizations with AWS science and strategy experts with deep experience in AI/ML and generative AI techniques. Public sector organizations can now use MAESTRO, to create cost-efficient generative AI tools tailored to agency-specific use cases. Within 9 months of launch, MAESTRO was adopted by 20 public sector organizations, corresponding to over 45 project teams and over 300 data scientists and ML engineers across Singapore.

MOM, the ministry responsible for the formulation and implementation of workforce policies in Singapore, used MAESTRO to drive quicker operations and policy interventions. It developed a generative AI sensemaker tool that uses clustering algorithms and an LLM to extract insights from documents. Within 3 months, MOM used the tool to process over 1 million documents, improve insights extraction by 60 percent, reduce sensemaking time by 50 percent, and save over 2,000 work hours.

MOM also enhanced job matching for employees and job seekers. It developed the Singapore Standard Occupational Classification (SSOC) Autocoder, which automatically sorted job postings based on occupational codes. Within 3 months following its launch, the SSOC Autocoder had processed 10 million job postings with 92 percent accuracy.

The Central Provident Fund Board (CPFB) of Singapore is a statutory board whose mission is to enable Singaporeans to have a secure retirement through lifelong income, healthcare financing, and home financing. CPFB used open-source, quantized Generative AI models deployed through MAESTRO to summarize call transcripts for the ~600,000 calls that it receives on average each year. These summaries are vital for follow-ups, investigations, and quickly identifying emerging issues.

GovTech MAESTRO incorporated generative AI capabilities using Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies, and Amazon SageMaker JumpStart, an ML hub with foundation models, built-in algorithms, and pre-built ML solutions. With ready-made models, public sector organizations can build and deploy AI solutions without investing significant time and resources in training models from scratch.

Furthermore, with quantization techniques, teams can deploy multiple smaller, right-sized models instead of a single large general model. Overall, by working with the AWS Generative AI Innovation Center, MAESTRO improved cost performance for generative AI workloads by up to 75 percent.

To simplify the creation, collaboration, and management of ML workflows, GovTech built MAESTRO using Amazon SageMaker Studio, a single web-based interface for end-to-end ML development, and Amazon SageMaker Canvas, a no-code ML interface. With a unified, web-based, no-code interface, MAESTRO reduces complexity and simplifies the building, training, and deploying of ML models even for non-technical staff. By lowering the barriers to entry, MAESTRO has enabled government agencies across the board to rapidly adopt and leverage cost-effective, production-ready Generative AI solutions, accelerating the implementation of various use cases.

"Leading organizations realize that scaling generative AI cost-effectively is critical. We're proud to collaborate with GovTech to make generative AI production-ready across Singapore's public sector,” said Nieves Garcia, AWS Generative AI Innovation Center Strategy Lead for Asia Pacific,  Japan and Greater China Region.

GovTech plans to deepen its partnership with AWS to expand generative AI and AI/ML operations and best practices to more government agencies. By optimizing for cost efficiency, GovTech aims to facilitate fiscally responsible generative AI development among public sector organizations.

“We're thrilled to partner with AWS to make generative AI more accessible and sustainable for our agencies. This allows us to harness generative AI's potential while responsibly managing resources,” said Jeffrey Chai, Product Manager of MAESTRO, Government Technology Agency.