peopleimages.com - stock.adobe.c
Grab opens AI centre of excellence, bets on home-grown AI
Southeast Asian super app Grab has launched an AI centre of excellence in Singapore, aiming to develop its own specialised foundation model, bolster developer productivity and create 50 new roles, among other goals
Southeast Asian super app Grab has opened an artificial intelligence (AI) centre of excellence at its Singapore headquarters to build AI capabilities that enhance accessibility, boost developer productivity and contribute to smart city initiatives across the region.
Speaking at the launch event, Singapore deputy prime minister and minister for trade and industry Gan Kim Yong noted that such centres will help grow a pool of AI talent and practitioners with the skills to create, implement and deploy AI systems. Grab’s AI centre of excellence is expected to create 50 high-value jobs across product, engineering, data science and analytics by 2025.
Suthen Thomas Paradatheth, Grab’s chief technology officer, elaborated on the company’s AI ambitions, noting that with over 1,000 AI models already powering the Grab platform, “we’ve seen firsthand how we can leverage AI to improve lives”.
“Recent advancements in AI hold the promise to transform how we interact with technology, foster innovation and bridge digital divides,” he added. “Grab is uniquely positioned to turn these breakthroughs into meaningful solutions that tackle real-world challenges by leveraging our rich data, expansive platform and passionate talent.”
A key undertaking for the new centre of excellence will be the development of Grab’s own specialised foundational model. Paradatheth explained this is not a large language model (LLM), but a sequence prediction model that leverages the same transformer architecture powering LLMs.
“What it predicts aren’t the next words, sounds or video frames, but rather, actions by users,” he said, adding that by predicting sequences of user interactions, Grab aims to improve user experience and deliver hyper-personalised recommendations.
Even as Grab develops its own foundation model, Paradatheth said the company will continue to partner with other model providers when appropriate, adopting a “horses for courses” strategy. This includes exploring regional alternatives, such as AI Singapore’s Sea-Lion model, and using established models from players like OpenAI for specific use cases.
Read more about AI in APAC
- Pharmaceutical industry experts at Gitex Asia 2025 discuss how AI is revolutionising drug discovery, from accelerating clinical trials and diagnosing rare diseases to validating traditional medicines.
- Telcos from Singapore, Malaysia and Indonesia are moving beyond the 5G hype as they focus on driving enterprise 5G and AI adoption to drive value for customers.
- Singapore researchers are using Google DeepMind’s AlphaFold to understand the intricate protein interactions at play in Parkinson's disease, paving the way for potential diagnostics and treatments.
- The Teraju AI Selangor initiative will provide state agencies with access to Google Cloud’s AI tools and training to build AI applications in areas such as education and healthcare.
For example, the centre of excellence is building a voice assistant using OpenAI models to help visually impaired users book rides via voice commands. Developed in collaboration with the Singapore Association of the Visually Handicapped, the assistant has been fine-tuned with 80,000 local voice samples, boosting its recognition of Singaporean accents and building names from 46% to 89%.
To further enhance its accuracy, Grab will launch a voice donation feature in its app in June 2025, enabling Singapore users to contribute voice samples to make the assistant more intuitive and inclusive.
Another focus for the centre is improving developer productivity. Through an in-house LLM kit that abstracts key components like security, Grab developers can speed up the development of generative AI applications from one and a half weeks to just half a day.
“The idea here is to get rid of things that are necessary but not immediately relevant to the problem we’re trying to solve,” Paradatheth told Computer Weekly. “For example, we want people to build securely, but we don’t want them to have the cognitive overhead of remembering every single detail related to security.”
Paradatheth noted that the LLM kit has already facilitated the creation of over 100 applications, including an AI assistant that provides merchants with business advice on boosting marketing and outreach efforts and adjusting pricing strategies.
Finally, the centre of excellence will also develop new smart city capabilities for Grab’s KartaDongle and KartaDashcam internet of things devices, going beyond their intended uses in mapmaking, fleet management and safety. These capabilities include real-time road hazard detection, traffic intelligence for urban planning and real-time flood monitoring as part of Grab’s partnership with PUB, Singapore’s national water agency.
Ultimately, Paradatheth said the centre would serve as an “accelerant”, enabling Grab to take bolder bets and focus on last-mile localisation critical for its diverse Southeast Asian markets. He also noted that Grab aspires to eventually “export these proudly Southeast Asian-first innovations to the global stage”.
