Super app Grab is no stranger to using artificial intelligence (AI) to deliver its services.
Its machine learning model platform, Catwalk, underpins more than 1,000 AI models in production to address the needs of its partners and users. With Catwalk, Grab provides users with real-time decision-making across its services and delivers personalised experiences like restaurant recommendations, loyalty rewards, and financial offerings tailored to users’ preferences.
Besides that, Grab has refined its last-mile guidance system for delivery partners by integrating large language models (LLMs) with point-of-interest data and historical customer notes. Doing so provides drivers with more precise drop-off instructions, enabling them to complete more trips every hour, leading to higher earnings, while expediting food delivery to consumers.
Amazon Web Services (AWS) is among the key enablers of Grab’s AI initiatives. Grab not only built Catwalk on Amazon Elastic Kubernetes Service (Amazon EKS) but also stores hundreds of petabytes of data and processes over 200 TB of data on AWS daily, which forms the foundation of Grab's advanced analytics, ML, and AI initiatives.
“We believe in constant innovation to drive better outcomes to our customers and partners, both drivers and merchants. So, there are three key dimensions of our tech strategy. One is cost optimisation. Second is being able to scale securely, swiftly and across all our markets and across our customer base. And third is how do we use data and AI to enhance the quality and experience of our products. We see AWS as a really good partner in enabling all aspects of our technology strategy,” says Mohan Krishnan, head of engineering, technical infrastructure for Grab, in a recent media briefing.
Supported by the cloud
See also: Unlocking the potential and navigating the risks of AI in the future of manufacturing
A scalable and robust IT infrastructure is crucial for driving constant innovation at Grab. Krishnan explains: “Innovation is built on the ability to experiment and keep iterating and improving our capabilities. To that end, we need to make sure our engineers and our product teams can focus on that aspect without being burdened with managing or scaling out the supporting tech infrastructure. AWS has been a key partner in abstracting a lot of that complexity so that our teams can focus on getting the best customer experience to our customers.”
Every second, Grab processes over a hundred transactions, receives over 500,000 GPS pings and fulfils over 50,000 ETA requests. It is able to do so by using Amazon Relational Database Service (Amazon RDS) as its transactional database coupled with Amazon DynamoDB. This ensures high availability, scalability, and adaptability of its platform to drive exceptional customer experience fuelled by more accurate searchable data.
AWS also empowers Grab to seamlessly adapt to evolving customer needs by dynamically adjusting resources in response to demand fluctuations. During peak times like holiday sales, Grab can easily accommodate increased traffic to make transactions across the super app seamless. Conversely, during off-peak periods, resources can be scaled down to save costs.
See also: Philips and Singapore General Hospital to jointly elevate MRI training across Asia Pacific
Besides that, AWS provides Grab with a stable and scalable infrastructure to support its continued expansion into Southeast Asia and new business streams. This includes rolling out the Advance Booking and Group Order feature across the region.
With AWS, Grab was also able to build and launch digital banks in Singapore and Indonesia, as well as GX Bank in Malaysia. Launched in under 16 months, GX Bank managed to scale rapidly to serve close to one million customers within the first year of launch.
Although the bulk of its workloads are running on AWS, Grab also utilises other cloud platforms like Azure and Google Cloud. “Running multi-cloud takes effort such as building the right skill set to ensure teams select the right cloud for their workloads and are able to work across multiple clouds. But we feel that a multi-cloud approach is beneficial as it allows us to use the best capabilities of different cloud providers. To simplify our multi-cloud set-up, we’ve made AWS our preferred cloud provider, where we’re running about 80% of our workloads on it,” shares Krishnan.
Optimising cost
As Grab seeks to balance growth with cost-efficiency, it relies on AWS Cloud to power most of its operations across Southeast Asia. This includes its mobility, delivery, and financial services verticals, as well as its digital banks in Indonesia, Malaysia, and Singapore.
For instance, it is leveraging AWS’s purpose-built databases. It has also migrated more than 400 back-end application services from traditional virtual servers to AWS Graviton2 processors to drive high performance, as well as cost and energy efficiency.
Furthermore, it uses custom-designed AWS Inferentia chips with specialised ML inference capabilities to cost-efficiently power its AI-powered services, including map enhancements and fraud detection in its digital banks.
“[To help Southeast Asian businesses like Grab thrive in the digital economy], it’s extremely important for AWS to continue providing the best tech stack – from infrastructure (including our Inferentia chips) to middle layer like Amazon Bedrock that provides a choice of LLMs, to services such as intelligent virtual assistant Amazon Q and Amazon Nova foundation models. We will also continue to invest in skills and workforce enablement to ensure we can address the demand for [AI talent],” says Gunish Chawla, managing director of Commercial Enterprise, Digital & SMB for Asean at AWS, in the same media briefing.
