Case Study – JARVIS
Embedding Predictive Intelligence into Live Rail Engineering
SMRT Trains · Singapore
How can operators anticipate rail failures before they disrupt service at scale, without compromising safety or engineering accountability?
Modern rail networks generate vast volumes of data from assets, condition monitoring systems, inspections, and maintenance records. Yet for many operators, this information remains fragmented across systems, spreadsheets, and departmental silos. Insights emerge late, troubleshooting relies heavily on experience, and failures are often addressed only after they affect service.
At SMRT, operating one of the world’s most intensively used metro networks, this challenge was acute. Engineering teams needed to make high consequence decisions under tight access windows, with zero tolerance for safety incidents or unplanned disruption. Existing systems could report historical performance, but offered limited foresight into emerging risks. Any solution had to improve anticipation, not just visibility and it had to work inside live operations where reliability and public trust are non negotiable.
The STRIDES Approach
Engineering Predictive Maintenance for Real Operations
STRIDES designed and delivered JARVIS, an AI enabled engineering intelligence system architected to support predictive maintenance and operational decision making in live rail environments.
Built on an Oracle enterprise data platform, JARVIS combines Oracle’s scalability, security, and data management capabilities with STRIDES’ engineering logic, operational workflows, and predictive algorithms. Rather than deploying analytics in isolation, STRIDES engineered JARVIS as an integrated decision support system aligned to how rail engineers actually work.
The system was purpose built around operational realities:
Multiple legacy data sources across maintenance and operations
Safety critical decision thresholds
Limited engineering access windows
The need for explainable, auditable recommendations
Key elements included:
JARVIS consolidates asset lifecycle data, maintenance history, condition monitoring readings, and operational context into a single trusted view, eliminating manual data correlation and information silos.
STRIDES designed predictive models to identify early degradation patterns and emerging fault risks, enabling engineers to intervene before failures escalate into service impacting events.
All insights are transparent and interpretable. JARVIS augments human judgement rather than replacing it, ensuring engineers retain accountability while benefiting from consistent, data driven recommendations.
A conversational interface allows engineers to interrogate complex datasets quickly, reducing reliance on specialised analytics skills and preserving institutional knowledge within the system.
Crucially, JARVIS was designed, tested, refined, and operated within SMRT’s live rail network, ensuring it delivered value under real operating pressure—not laboratory conditions.
Impact
Earlier identification of emerging asset failure risks
Shift from reactive fault response to predictive intervention
Faster, more consistent engineering decision‑making
Improved maintenance prioritisation based on risk and condition
Engineering knowledge captured and shared through the platform
Why It Matters
Predictive Maintenance Only Works When Engineers Trust the System
JARVIS shows how execution‑driven engineering intelligence, designed by operators and proven in live service, can deliver outcomes that theoretical analytics and stand‑alone platforms cannot.
What Other Operators Can Learn
In safety critical rail operations, analytics deliver value only when insights are trusted, explainable, and grounded in operational reality. JARVIS demonstrates that predictive maintenance is not a software challenge, but an engineering one—requiring deep understanding of how systems behave in service.
Multiple legacy data sources across maintenance and operations
Safety critical decision thresholds
Limited engineering access windows
The need for explainable, auditable recommendations
Enterprise platform powered by Oracle | System architecture, engineering logic, and operational design by STRIDES