What it takes to engineer an ‘uneventful’ train ride for commuters

Parts of Ms Celine Tan’s usual train journeys on the Circle Line looked quite different earlier this year – and took about up to 10 minutes longer.

At MRT stations such as Mountbatten, Dakota and Paya Lebar, staff wielding light sticks guided commuters along designated routes on packed platforms. The crowd-control measures were introduced during tunnel strengthening works that, from Jan 17 to April 19, 2026, added up to 30 minutes to some peak-hour journeys along the Circle Line.

It was an inconvenience, admits Ms Tan, 30, who had to factor extra travel time. “I don’t like it, but I understand that this was necessary,” says the operations executive, who recalls reading about the tunnel strengthening works in the news.

Ms Tan takes the MRT three to four times a week to meet friends on weekday evenings and weekends. She cycles to her workplace, which is 15 minutes away from home.

The Circle Line tunnel strengthening works provided a glimpse into the complexity behind a rail network which Ms Tan rarely thinks about. She assumes “a lot of coordination and technology must be happening behind the scenes to keep (such a large system) running smoothly every day”.

Like Ms Tan, 70 per cent of commuters polled say backend technologies used in public transport matter to them, even though they largely operate out of sight, revealed a recent survey.

The survey, commissioned by SPH Media and SMRT, polled 1,010 public transport users in Singapore in February 2026. Its aim: to understand how Singapore commuters perceive the use of advanced technologies like AI and automation in public transport, and their concerns and expectations.

Respondents were sourced from the Kantar Profiles Audience Network. Kantar is an independent market research company.

Seventy-five per cent of respondents also said knowing about these backend technologies gave them greater confidence in the public transport system’s safety and performance.

For Ms Tan, the awareness helps her “better appreciate how complex train operations actually are” – especially when public transport is the “cheapest and most convenient way” for her to get around. 

“It also makes me more understanding when disruptions happen occasionally because there are so many moving parts involved.”

Behind the scenes

But most commuters expect their train journeys to remain simple. For Ms Tan, this means “trains that arrive on time, and short waiting times”. 

And when disruptions do occur, she wants “clear announcements about what to expect”.

Delivering that seemingly straightforward expectation requires what Mr Albert Soh, head of Engineering Analytics at SMRT’s innovation arm Strides Technologies, describes as “invisible teamwork” between technology, people, and processes. Here’s what that looks like.

The public transport operator is increasingly using technologies like AI and automation to detect potential faults early and prevent them from disrupting commuter journeys. 

One example is Jarvis, rolled out in January 2026. The intelligent data platform enables SMRT engineers to shift from post-mortem investigations towards pre-emptive analytics, says Mr Albert Soh, head of Engineering Analytics at SMRT’s innovation arm Strides Technologies.

Jarvis is developed by Strides Technologies and built on cloud and AI infrastructure by software company Oracle.

In the past, engineers had to manually study operational data to analyse faults and support service recovery, explains Mr Soh, who led the development of the Jarvis platform. 

Today, Jarvis can process large volumes of maintenance and operational data to detect subtle signs of degradation. This enables engineers to fix issues before faults happen. 

The aim is to help engineers identify risks earlier, prioritise interventions and fix potential faults before they occur, says Mr Soh. But he stresses that predictive AI is “not a holy grail”. 

“We can’t predict every single failure mode,” he says, referring to the different ways equipment or systems can fail. For example, electrical parts can fail suddenly without showing early signs of degradation, he explains.

Mr Soh adds that sensors within train systems generate massive amounts of data. The aim of Jarvis is to put these data to use more efficiently – and try to reduce disruption as much as possible. 

Even as artificial intelligence takes on a greater role in helping to spot potential issues, people remain central to the rail system, says Mr Soh.

“We want to make sure that our people are well-trained and well-equipped (to make full use of the AI tools),” he says, adding that AI can make jobs easier for existing employees, and “accelerate learning and training for new (hires)”. 

One way SMRT is doing so is through Jarvis’ generative AI capabilities. 

The platform is powered by a large language model trained only on SMRT’s internal data and engineering documents. It is not connected to the internet, Mr Soh says.

Previously, when there is a train or track fault, staff had to sieve through hardcopy manuals to identify the right procedures for each scenario.

Now, engineers can ask Jarvis questions through a chatbot-style interface, Mr Soh explains. The system retrieves relevant procedures, compares different scenarios and advises staff on next steps.

Mr Soh describes Jarvis as a “single source of truth” that helps staff to make better operational decisions faster. The goal: to improve safety and reliability of the train system overall.

But before AI systems can be layered onto operations, it is important to first build “good and robust processes”, Mr Soh says.

These processes span what he describes as a sequence of “prediction, detection, and response”. 

Predicting a fault is only one part of the equation. When something goes wrong, says Mr Soh, engineers must be able to pinpoint issues accurately, communicate the information quickly and coordinate responses across teams.

The challenge is compounded by the growing complexity of modern rail systems, which Mr Soh describes as a “system of systems”. It refers to an interconnected network of signalling, communications, power, trains, track and station operations.

When one of these systems has issues, it can result in wider impact across the network if not managed quickly, adds SMRT group chief engineering officer Ang Hang Guan.

For example, a stalled train affects every train behind it on the same line.

When such incidents happen, Mr Ang says, operations teams must work quickly to get the affected train moving. Services also need to be rerouted so that only a segment of the network is disrupted while the rest of the rail system continues running.

Despite the scale of SMRT’s rail system being close to 200km, Mr Soh acknowledges that for commuters, the ideal outcome is far simpler. A smooth, uneventful ride shows that things are working well, he says. 

“With the right foundations – people, processes and technology – in place, the outcome can then be delivered systematically and consistently,” Mr Soh says.

In partnership with SMRT