SMRT's JARVIS AI Platform Helped Trigger 500+ Rail Maintenance Inspections in Its First Year
Source: SMRT
Singapore's rail operator is quietly becoming one of the more interesting case studies in how a legacy infrastructure business can industrialise AI. For years, MRT maintenance ran on scheduled checks and engineer intuition — but with a network that now carries more than two million passenger journeys a day,...
Singapore's rail operator is quietly becoming one of the more interesting case studies in how a legacy infrastructure business can industrialise AI. For years, MRT maintenance ran on scheduled checks and engineer intuition — but with a network that now carries more than two million passenger journeys a day, SMRT has been building JARVIS, an internal data-and-AI platform, to shift the model from "fix when it breaks" to "predict before it fails".
The numbers from SMRT Trains' FY25/26 results, released on August 5, show the platform is earning its keep. JARVIS — developed by STRIDES Technologies with Oracle, and launched in phases from January 2026 — has so far initiated more than 500 maintenance inspections after flagging early signs of component degradation. Phase 1 integrated data from the North-South and East-West Lines with 11 analytics modules covering assets like platform screen doors and track-side radio equipment; Phase 2 extends the platform from rolling stock and signalling into track and power systems.
The results also give the AI push a financial frame. SMRT Trains recorded EBIT of S$16.2 million in FY25/26, up from S$4.9 million a year earlier, with revenue rising 5.6 per cent on higher ridership. Its lines hit an average Mean Kilometres Between Failures of more than two million train-km from May to July — double Singapore's one-million benchmark and in line with some of the world's best metros. JARVIS is part of a broader engineering drive that also includes Depot 4.0 upgrades at Mandai Depot and a joint rail innovation lab with Guangzhou Metro opened in June.
The interesting part is what JARVIS represents for the wider Singapore ecosystem. Most public-sector AI conversations centre on chatbots and policy; here, an operator is using predictive analytics and large language models on maintenance data to make decisions about physical infrastructure worth billions. That is exactly the kind of applied AI Singapore's Smart Nation push wants more of — and it sets a template for other asset-heavy industries, from utilities to aviation, that are still collecting data without acting on it.
Why it matters for Singapore: Rail reliability is a daily, tangible measure of whether AI delivers. Every train-km gained through predictive maintenance is public proof that the technology pays for itself in a critical national service — and with the Cross Island Line on the horizon and the network only getting bigger, SMRT's bet on AI-enabled engineering is really a bet on how Singapore runs its infrastructure for the next two decades.
