MEPX AI Solutions is an early-stage AI research practice exploring what becomes possible when artificial intelligence is applied to the full coordination challenge of building design — from first intent to every discipline, simultaneously. Our approach is MEP-led: we believe the discipline most often treated as an afterthought should be the one that drives the entire design from the start. It is time to pay attention to the elephant in the room.
Building design has always required the simultaneous resolution of multiple disciplines — interior, MEP services, architecture, and structure. In practice, these disciplines are resolved sequentially, by different teams, with coordination treated as an afterthought. The result is conflict, waste, and delay.
We believe this is a solvable problem. Not by better software for each discipline in isolation, but by rethinking the design process as a single, coordinated act — where every discipline is informed by, and responds to, every other from the very first decision.
That is what we are building toward. We are at the beginning.
Our work is focused on modular and precast construction — a domain where design coordination is both most consequential and most tractable to automate. The constructability logic of precast demands precision in every discipline before a single element is manufactured. That precision, we believe, should be generated — not manually coordinated.
We are not yet a service. We are a research and development endeavour, registered in Western Australia, working on a system that does not yet exist at the scale we intend.
MEP is among the most deeply interconnected disciplines in building design — woven into the fabric of every floor, wall, and ceiling. It intersects with structural elements, is constrained by architectural decisions, and is driven by interior requirements simultaneously. The coordination surface it presents across all other disciplines is, in most building types, broader and more consequential than any other single service.
When MEP coordination fails — and in conventional practice, it routinely does — the consequences are not cosmetic. Clashes discovered on site mean stopped works, redesign under time pressure, and cost overruns that compound across every affected trade. In multi-residential construction, where the same coordination decisions repeat across dozens or hundreds of units, the cost of getting it wrong early is multiplied accordingly.
This is why our approach begins with MEP — not because it supersedes structural complexity or architectural intent, but because resolving it early, against every other discipline simultaneously, removes the most consequential source of failure before construction begins.
Every building requires the resolution of six interdependent design disciplines. The gaps between them — in time, in teams, in tools — is where most coordination failure originates. Our work is focused on what it would mean to resolve all six at once, from a single design intent.
Space, layout, and occupancy requirements. The origin of all downstream decisions.
Thermal comfort, ventilation, and air distribution — derived from how the space is used.
Power, data, and lighting — shaped by occupancy, furniture, and zone function.
Water supply and drainage — constrained by layout, coordinated with all other services.
Envelope, openings, and spatial character — derived from what is inside, not the reverse.
Structural system and constructability — resolved against the complete, coordinated design.
In conventional practice, these disciplines are addressed by separate teams across weeks or months. Coordination conflicts are discovered late. We are working on a different approach.
SAMI — our design philosophySAMI is the organising principle that guides our approach to building design coordination. It defines the four primary design disciplines and the dependency between them — each one a response to what precedes it.
Structure responds to what the space requires — not the other way around. The structural system is a consequence of design intent, not a constraint imposed upon it.
The architectural envelope follows what is inside it. Facade, openings, and spatial character emerge from interior and services — resolved, not assumed.
Mechanical, electrical, and plumbing services are married to the interior from the moment the space is defined — an integral part of the design, not a layer added to it.
Interior design is where every decision begins. The requirements of the occupant define the space. Every other discipline exists to serve that definition.
A process that resolves all disciplines simultaneously can produce a fully coordinated building design in a fraction of conventional time. But that speed is only valuable if the construction method can match it.
Conventional construction cannot. It is sequential by nature — trades follow each other on site, conflicts are discovered physically, and changes are expensive. The faster the design, the more painfully the construction bottleneck shows.
Modular and precast construction can. Manufactured off-site to precise specifications, it demands exactly the kind of fully coordinated, conflict-free design our system is built to produce. For multi-residential development — where design decisions repeat across many units and construction tolerances are tight — the pairing is not incidental.
Precise automated design requires a construction method precise enough to take advantage of it. Modular and precast is that method. The two were made for each other.
Modular and precast construction reduces construction waste by design — components are manufactured off-site to precise tolerances, minimising material over-ordering, cutting and site waste that characterise conventional construction.
MEPX designs to performance targets from the first configured input. Energy load analysis, thermal comfort modelling, and compliance with the National Construction Code are not post-design checks — they are embedded in the coordination logic from the start.
For multi-residential buildings in Western Australia, this means every unit configuration is assessed for thermal performance under the NatHERS framework — Australia's residential energy rating standard — as part of the automated design process, not as a separate consultant engagement after the design is done.
Our energy modelling capability is built on OpenStudio and EnergyPlus — the leading open-source building energy simulation engines used by research institutions, governments, and design practices globally. These tools allow whole-building energy analysis at a level of detail that is currently inaccessible to most residential projects due to cost and time constraints.
Automated design changes this equation. When the design is generated rather than drawn, energy simulation can run at every iteration — not once at the end. The result is a building that is optimised for performance from its origin, not retrofitted for compliance at its completion.
The tools we build on are deliberate choices — enterprise cloud infrastructure, large language models for reasoning and code generation, Revit for BIM authoring, Revizto for cloud-based coordination and clash detection — replacing desktop-dependent workflows with live simultaneous access across all disciplines — IFC as the open interoperability standard, and Australian Standards as the compliance foundation. Subscription dependency is kept to a minimum by design.
What drives them is not the tools themselves. It is the domain knowledge encoded into the system — forty years of active MEP engineering practice, applied across complex building projects, accumulated through exactly the coordination failures this system is designed to prevent. That knowledge defines the rules, the dependencies, the sequencing, and what a correct, buildable, coordinated outcome actually looks like.
The role of AI here is precise: large language models act as the technical translator between deep domain expertise and working implementation. The engineer defines what good coordination looks like. The LLM writes the code that executes it. The infrastructure deploys it at scale. No single layer is sufficient without the others — and the layer that cannot be replicated is the one that took forty years to develop, and is still growing.
In conventional practice, a designer interprets a client brief and produces drawings to satisfy it. MEPX inverts this entirely. The project requirement — unit mix, occupancy types, zone configurations, and jurisdiction — is entered directly into a front-end configurator. Each zone is defined by type and scale: a sleeping zone, a kitchen, a bathroom, a study — each sized according to a T-shirt scale from XS to XL, reflecting the real range of human need and budget. Medium is the standard. Everything else is a deliberate choice.
Jurisdiction is not a manual checking process — it is a configured input. Western Australia is our pilot jurisdiction, with its specific codes, standards, and authority requirements fully embedded in the system. Additional jurisdictions follow as the system matures. The requirement drives the drawing — not the other way around.
The backend of MEPX is not a single AI. It is an agentic organisation — a coordinated system of dedicated AI agents, each expert in its own discipline: structural, architectural, MEP, and interior. Each operates within its domain. Each coordinates with the others according to the SAMI dependency hierarchy — the same logic that governs the design philosophy governs the system architecture. This is SAMI not just as a principle — but as a working organisational model, implemented in code.
The system produces two categories of output: individual discipline drawings for each agent domain, and a fully coordinated drawing set — all disciplines resolved simultaneously against each other. Every individual discipline drawing is validated and signed off by a qualified human professional in that field. The coordinated MEP drawing is signed off by a qualified MEP coordinator. The complete coordinated SAMI drawing is signed off by a SAMI coordinator — the human who takes professional responsibility for the whole. AI generates. Qualified humans validate. That is not a limitation of the system — it is the correct and only professionally defensible model for building design. It is how we intend to operate.
We are based in Perth, Western Australia. Our initial focus is on the Australian market — where housing supply constraints and the case for industrialised construction are both acute. Our design framework is built to support multiple jurisdictions as our work matures.
Connect — the conversation is openWestern Australia is our pilot jurisdiction. Local codes, standards, and authority requirements are fully embedded in the system from day one.
Our initial focus is multi-residential modular and precast development — the domain where automated design coordination delivers its greatest advantage.
The T-shirt sizing principle means a single coordinated design logic scales from a single studio to an entire building — and from one jurisdiction to many.
If you are working in housing delivery, industrialised construction, building technology, or related investment — we are interested in your perspective. We are not yet offering services. We are building, and we are listening.