, Singapore’s Whole-Of-Government data exploitation platform, provides advanced analytics and machine learning capabilities with cloud

By Amazon Web Services (AG), built by the Government Technology Agency of Singapore (GovTech Singapore)’s Data Science & AI Division, has partnered with Amazon Web Services to provide advanced analytics and Machine Learning Operations (MLOps) capabilities to government agencies. GovInsider speaks to Jeffrey Chai, Product Manager at GovTech, to learn more.

The (AG) team works on a central data exploitation platform to democratise data analytics and machine learning tools across government agencies. Image: GovInsider

According to Forbes, data science teams are moving at unparallel speeds – machine learning problems are becoming more complex and models have to be leveraged at scale for any organisation to derive actionable insights. These teams will need to have access to state-of-the-art analytics tools, coupled with powerful computing resources and means for workflow automation.


These are just some of the key reasons why GovTech Singapore expanded its central data exploitation platform to the Government on Commercial Cloud (GCC) in 2022, says Jeffrey Chai, Product Manager at GovTech Singapore’s Data Science and AI Division (DSAID). is now being used by more than 1,600 users across 80 government agencies, enabling data science practitioners across the Singapore Government to tap on a variety of data analytics tools and scalable compute power, Jeffrey shared at the AWS ASEAN Summit that took place on May 2023.


Notable agencies onboarded onto AG to-date include the Ministry of Manpower, Ministry of Foreign Affairs, Housing Development Board and SkillsFuture Singapore, says Jeffrey.

GovInsider speaks to Jeffrey to learn more.

Democratising analytics can be as easy as ABC


Jeffrey shares that AG was first developed as a secure platform in 2020 to better support data science teams performing advanced analysis and machine learning on government data, and in compliance with government architecture and security requirements.


The vision for AG is to be that readily available environment for any agency to work on their data projects right away without the need for them to expend additional time and effort to develop equivalent systems on their own.


“With a centralised platform like AG, the management and maintenance is kept consistent. There are also savings accumulated from leveraging a central service and reducing duplication of efforts,” he says.


According to Jeffrey, it delivers the “ABC” of “access to up-to-date analytics tools and code libraries, better compute resources from the platform itself, and most importantly, seamless collaboration across government agencies”.


“Through the common platform, users are able to share codes and collaborate with anyone and everyone within the same agency or even across the other agencies,” he explains.


Government data scientists currently tap AG for a diverse range of use cases such as policymaking, service delivery and internal operations, he shares.


These include analysing, predicting and visualising customer feedback data, development of chatbots, report summarisation, ML-assisted case outlier detection, automated data processing for governance checks on grants, campaign planning and workflow processes.

Standardising ML deployment across government


Now, the team is focused on accelerating machine learning innovation by empowering the Whole of Government with machine learning operations (MLOps) capabilities.


MLOps refers to the process of streamlining the lifecycle of machine learning models, from training models to production and monitoring, explains Karthik Murugan, Head of AWS Analytics Services. This is key to scaling machine learning algorithms in complex organisations such as governments, he says.


“The standardised workflows in place will help data science teams streamline their processes to ensure quick and efficient model deployment. This will empower data teams both within and across government agencies to collaborate and accelerate machine learning innovation through real time co-working capabilities.”


“While there is growing interest to embrace artificial intelligence and machine learning over the recent years, manual processes can slow down the rate of adoption,” Jeffrey says.


The need for MLOps adoption across government agencies was also highlighted in the recent presentation by GovTech Singapore during Public Sector Day Singapore – “Building a Secure MLOps framework”.


Speakers shared that the success of the Proof-of-Concept with SkillsFuture Singapore demonstrated why MLOps is the way forward for organisations to truly exploit their investments in AI/ML technologies, as no model is perfect at birth and any good ML model can go rogue anytime without monitoring and governance. What they lack is an end-to-end platform to bring their Proof-of-Concept to production.


This is why AG is critical as the go-to platform that can empower government agencies to automate their ML workflows and scale machine learning projects end-to-end, from data ingestion to model and application endpoints in production, Jeffrey says.

Tapping on cloud services for MLOps


Expanding to GCC was a no-brainer for the team, which sought to anticipate future demand from government agencies for more compute resources (including GPUs) and higher elasticity of these resources, he explains.


“Instead of desktop-based or data-centre-based platforms, AG on GCC is the game changer that every data science team needs for achieving greater productivity and efficiency as they are able to develop and deploy ML models at scale through automation and standardisation,” says Jeffrey.


On top of cloud-native services, the team also carried out additional infrastructural development efforts to make sure that the necessary security guardrails, corresponding to compliance requirements, were put in place for AG so that users can have better peace of mind when using these services on the platform, he explains.


Jeffrey also shares that while his team had trailblazed MLOps capabilities on AG via Amazon Web Services (AWS) SageMaker, it is essential for the central platform to also build similar capabilities in multiple cloud environments to cater to agencies that are using other cloud service providers.

Positioning Machine Learning with Amazon Sagemaker


“With Amazon Web Services (AWS) Sagemaker, AG has provided government agencies with the means to perform advanced analytics and machine learning operations in an automated and scalable manner, through GCC,” says Jeffrey.


“Through close collaboration, we worked with AWS to study ML use cases from potential user agencies and jointly bring in the cloud-native services which best support AI/ML technologies. The services include auto ML, ML pipelines, no-code ML, API services for model deployment and ready access to generative AI models,” he explains.


“Amazon SageMaker’s managed features allow customers to standardise workflows, create operational efficiencies, and ensure governance at scale,” says Murugan.


Murugan also shares that the AG team provides its user agencies access to Amazon SageMaker tools such as SageMaker Jumpstart, which gives users access to foundation models such as Large Language Models (LLMs) to build and put into production quickly.


Additionally, Amazon SageMaker Canvas offers a no-code interface that allows anyone to create ML models in minutes without any previous experience using interactive visual interfaces and point and click tools.


The MLOps services within AG come equipped with monitoring and governance mechanisms for models that have been productionised and deployed, he adds. Through SageMaker Model Monitor, ML engineers can monitor the performance of their machine learning models against four key indicators: data quality, model quality, model bias drift, and explainability.

Building a robust AI/ML community


Jeffrey and his team also worked closely with AWS to conduct knowledge sharing webinars and hands-on workshops for prospective agencies’ users, with the aim of wanting to build a stronger community of AI/ML practitioners across the Government.


As a strong advocate for collaborative learning, he believes that a community provides a channel for knowledge exchange, fostering diverse perspectives and ideas. It enables the pooling of resources and expertise, facilitating quicker progress to solve complex problems and drive innovation.


“Moreover, ethical and responsible AI development demands a collective effort to ensure fairness, transparency, and accountability. In essence, a robust AI/ML community is not merely a luxury but a necessity for the responsible and effective advancement of these technologies,” says Jeffrey.


Click here to find out more about on the Singapore Government Development Portal.