Muhammad Izzul Hakimi
Muhammad Izzul Hakimi bin Roslan was one of five engineers celebrated as digital and automation innovators in this year’s CIBSE Global YEN 30 under 30 Awards.
The category recognises young professionals driving innovation through digital technologies, automation and smarter approaches to building performance.
In his role as an inspection officer at CAMS Assethub, Roslan specialises in digital asset management and data-driven infrastructure decision-making.
His work aims to shift infrastructure management from reactive to proactive approaches, and his cross-disciplinary approach – combining civil engineering, digital systems and research – positions him at the forefront of intelligent infrastructure and resilient asset management.
Roslan recently graduated with first-class honours from RMIT University, in Melbourne, Australia, with a Bachelor of Engineering (Civil and Infrastructure). Alongside his industry role, he is undertaking a PhD at RMIT University, focused on developing a digital twin framework that integrates inspection data into predictive, evidence-based asset management systems.
At CAMS Assethub, he contributes to large-scale building condition audits, capturing and structuring asset data to support life-cycle planning and maintenance prioritisation.
In our Q&A below, Roslan explains that closing the gap between building design and real-world performance is as dependent on data disciplines as it is clever algorithms. With structured, contextual digital twins, he says asset managers can build trust in predictive maintenance and drive true operational efficiency.
Failures rarely stay contained to one component, so systems should be designed so those dependencies are visible, not buried in someone’s head.
With the current industry focus on net zero and operational carbon, how can better asset data and digital twins be used to tackle the performance gap in existing buildings?
The performance gap is mainly a data gap. Decisions on replacement, retrofits and maintenance are usually based on out-of-date condition data, meaning actual building performance continuously drifts away from its original design performance targets.
A digital twin can bridge that gap if it is continuously in sync with the real building and applied across the whole portfolio, instead of a single building.
How do you ensure the data captured during a building condition audit translates into a usable digital model?
We have to be disciplined when capturing such data. On the programmes I have worked on, for example, every asset is recorded against a fixed classification and rating scale using our own proprietary CAMS mobile app. Afterwards, it goes through a structured pipeline: export, an initial QA, format conversion, a second QA, then delivery.
This allows us to provide proper defect descriptions, which can be fed into our clients’ work-order systems and help with their own planning and renewals.
Your PhD focuses on a digital twin framework for predictive maintenance. What is the biggest hurdle in getting building owners to trust predictive data?
I am only a few weeks into my PhD, so this is early thinking rather than a finding, but my instinct is that it’s not about the algorithm; it’s that building owners can’t see what’s behind the number. I saw this at the Conference on Railway Excellence, by the Railway Technical Society of Australasia, which I attended in early August as an Emerging Rail Professional Scholarship recipient.
One paper talked about how the same AI models scored 98-100% accuracy on data queries when paired with a semantic context layer, whereas it was 84-90% with no context at all. The lesson generalises beyond rail: a system that is traceable in context and reasoning gives trust and accuracy together.
If you could mandate one change to how handover data is delivered at the end of a construction project, to make life easier for asset managers, what would it be?
Implementing structured, machine-readable asset records, with an accurate day-one condition baseline, would allow seamless integration directly into asset management systems, replacing static PDF or Excel handovers.
Current deterioration models often default to assuming brand-new assets are flawless simply because of a lack of better baseline data. Establishing a proper baseline at handover would fix all of that at the source, instead of asset managers figuring it out years later.
What is the most common roadblock you encounter when trying to capture clean, structured data from older building systems?
Older systems were never built to export anything clean. Sensors can be uncalibrated or undocumented, every legacy register uses its own tags and scales, and there’s often no digital baseline at all. Most often, not transferring data from legacy systems means the asset data fields can be all over the place and are not easily mapped to our own systems.
I have done floorplan capture using an iPad-based scanning app because older council buildings, more often than not, don’t have digital floorplans; they only have hand-drawn sketches or incomplete/not updated drawings.
Which digital tools, apps or capture methods have made your life easier on site, and which ones turned out to be more trouble than they were worth?
Structured mobile inspection apps with immediate sync have made the biggest difference; they kill the lag between capturing something on site and it existing in the system. That is what our consultancy, CAMS Assethub, essentially tries to provide to our clients: our own mobile inspection app named CAMS (Central Asset Management System).
It keeps an online database of all the asset registers of a portfolio from the data that was captured on site, which can then be used for deterioration modelling and financial planning. On the other hand, iPad-based room-scanning apps for floorplans produce output that looks good on screen but needs cleanup afterwards.
In your experience, where is the biggest disconnect between the physical reality of the assets you inspect on site and the digital models or asset registers the clients think they have?
Registers describe what was installed, not what’s actually there now. During my time auditing buildings at RMIT and at CAMS, I achieved a spatial variance – the physical to drawn area gap – of 1.05% and a 20% increase between planned and audited areas, well within the industry plus-minus 10% standard.
That shows how far a register can drift when it isn’t actively field-verified. On site, you constantly find modifications that were never logged, assets replaced without an update or entire small systems missing from the record.
How has your hands-on experience identifying failing or inefficient plant equipment shaped your view of how we should design new systems?
It has taught me to design for maintainers rather than just handover day. That requires establishing a clear day-one condition baseline, calibrating and documenting sensors instead of relying on black boxes that are bolted on and often forgotten, and applying a consistent classification scheme from the start.
Failures rarely stay contained to one component, so systems should be designed so those dependencies are visible, not buried in someone’s head.
What are we designing today that is going to be a nightmare for someone to audit or maintain in 15 years?
Highly instrumented, tightly controlled buildings with proprietary systems, no calibration plan and no structured way of keeping their data current. A sensor quietly drifting out of calibration, or a digital twin built once for handover day but never updated, look fine until someone actually relies on them – resulting in someone, in 15 years’ time, inheriting a system they can’t audit and can’t afford to remodel.