Use Cases

To-date, MAESTRO serves the AI/ML needs of more than 600 users across 40 public agencies, supporting a diverse range of use cases from enhancing service delivery to improving AI/ML productivity. Agencies are leveraging the platform to productionise their AI/ML models efficiently, allowing them to continuously develop, test and deploy models through the automation of ML and CI/CD pipelines. As a readily available AI/ML Ops environment built in compliance with government architecture and security standards, MAESTRO allows government agencies can work on AI/ML projects securely right away without having to expend additional time and resources to develop equivalent systems on their own.

MAESTRO enabled the Housing Development Board (HDB) to adopt a full MLOps workflow on GCC 2.0 via AWS SageMaker, together with the R Model Template developed in-house by the MAESTRO engineering team, providing an end-to-end solution where the model is developed and maintained within our SageMaker environment but readily consumed externally in the government network via API Gateway configurations. Such automation has led to time savings of approximately 26 man-days (0.5 man-day per week, 52 weeks in a year) annually from not having to manually re-train the models. In addition to cost avoidance from not having to develop an equivalent system, HDB also saves a total of 52 man-days per year on preventive maintenance (~12 man-days) and annual upgrade cycles (~40 man-days), which they previously required for their on-premise infrastructure.

MAESTRO enabled SkillsFuture Singapore (SSG) to productionise their key ML use cases seamlessly by providing them with a secure, enterprise-ready MLOps environment. SSG avoided significant developmental costs by not having to set up and maintain their own infrastructure. Instead, by leveraging MAESTRO's environment, they were able to accelerate the productionisation of key use cases like their Fraud Assessment Model and their Skills Extraction Algorithm. In addition, SSG was able to implement MLOps best practices and processes on MAESTRO, resulting in faster development iterations (~8 hours saved per release) as well as consistent deployments and real-time monitoring for greater model reliability and performance.

MAESTRO enabled the Ministry of Manpower (MOM) to accelerate and scale their LLM use case development that was previously constrained by the limited functionality and compute resources of their on-premise set up. Tapping on MAESTRO's connectivity with Container Stack via the Government Enterprise Network, MOM was able to deploy LLMs for several key use cases on GCC2.0, including their SSOC Autocoder and Sensemaker. The former has seen an improvement in the accuracy of job advertisement labeling (from 38% to 92%), while the latter has resulted in a 50% reduction in the time needed for textual sense making.