Commercial real estate and hotels
Chiller plants tuned for energy savings in malls and premium hotels. The same approach turns complaints into drafted maintenance actions, with a manager's review as the only manual step.
Systems that observe, reason, and intervene.
Software that runs daily operations in malls, factories, farms, clinics, kitchens and shops: a machine that asks for repair before it breaks, stock that reorders before the weekend, a consult written up before the patient leaves.
Every conversation is confidential. We typically reply within a day.
One pattern. Every operation.
-
Meters, sensors, cameras, documents, logbooks: the operation is already generating the evidence. The system takes it in whatever shape it arrives, including the creased and handwritten.
-
Your engineering rules, your prices, your history, your thresholds. Drift gets explained rather than merely flagged, and the explanation is one your own people can argue with.
-
Inside limits agreed beforehand it acts on its own: a setting retuned, an order raised, a record filed. Outside them, or unsure, it goes to a person and waits.
-
A server inside your network, not someone else's cloud. Data never leaves your firewall, and the system keeps running when the internet does not.
Most enterprise platforms are merely observation tools. Ours are architecturally different: agents that monitor, reason and act, across physical and digital operations alike.
Same seven flows. Different uniforms.
Kopitiam or factory: the uniforms differ, the flows do not.
- Salesopportunity, contact, quote
- Quote to orderthe deal gets signed
- Order to cashinvoice, collection
- Fulfilmentplan, procure, make, store, deliver
- Finance and investmentwhere the money sleeps
- Product lifecycleidea, production, sunset
- R&D and softwarethe development cycle itself
Bring us the flow that hurts: that is where a case study starts.
Built for the places where a mistake costs real money.
Some run in production today. The others are complete designs, drawn for the kitchen, the clinic and the warehouse. And the door is open for the problem that is none of these.
Agriculture and palm oil
Water, soil, weather, plant health: the daily settings that decide an estate's yield, watched everywhere at once. Born from research mapping 100,000 forest canopies from the air.
Factories and heavy industry
Old machines, new instincts. Production equipment watched for mechanical degradation, with maintenance orchestrated before downtime turns catastrophic.
Trading and distribution
Invoices, delivery orders, customs forms: paper someone retypes into your accounting or stock system. The records system reads, checks and files it instead.
Healthcare and clinics
A doctor loses the first minutes of every consult to the file and the evening to writing it up. This design returns both, and records never leave the hospital.
Food and beverage
Rain on a Saturday means thirty percent fewer walk-ins, and the ordering should know first. Stock from the weather, waste priced, menu costing live, reviews answered in your voice.
Logistics
The same lorries, more drops. Routes planned from the day's real orders, ETAs that are actually true, claims settled from photos, and the cold chain watched every degree.
Retail
The camera above the aisle already sees it. Shelf gaps raised within the hour, shrinkage patterned by when and where, and reorder points that move with the season.
What a client brings us is never published. Your brief stays yours; the designs here are our own.
Not a system you need built, but a room you need taught?
Conference talks, in-house training and hands-on workshops, for associations, teams and events. Every session is rebuilt around the room.
Cambridge science. Enterprise-scale engineering.
QuarkX founder Koay Xian Jing trained at Cambridge, where his published research mapped over 100,000 tropical tree crowns with deep learning.
Then enterprise scale: DPU offloading that doubled Storage Area Network IOPS at a global telecommunications giant, and agentic LLM workflows at a world-leading fintech that removed thousands of hours of manual work and earned CTO-level deployment approval.
He does not work alone. Where a system touches physical equipment, every control action is checked against the working rules of an engineering practice with more than twenty-five years in plant rooms.
QuarkX puts that profile, scientific depth and bare-metal engineering discipline, into autonomous systems for the physical world. Not prototypes: systems that run.
University of Cambridge
Natural Sciences · Computational Chemistry & ML
Regional Digital Platforms Group
Data Engineer · Digital Platforms
Global Telecommunications Giant
DPU Offloading Architecture
Global Fintech Leader
Lead Engineer · Agentic Systems
QuarkX Sdn. Bhd.
Founder · AI Systems Architect
Peer-reviewed work that shapes how we engineer.
Wiley Top Viewed Article
Accurate delineation of individual tree crowns in tropical forests from aerial RGB imagery using Mask R-CNN
Remote Sensing in Ecology and Conservation · Vol. 9, Issue 5
Asked by operators. Answered by engineers.
New to this? What AI can honestly do for an operation here, and what it cannot.
What does QuarkX actually do?
Traditional AI stops at a dashboard recommendation and waits for a busy person to act. Our systems close the loop: they observe, reason, and act themselves, inside written limits your team sets.
What is an "agentic AI" system?
A normal AI tool answers when asked. An agentic system is given a job and does it: watching, deciding, acting, checking. The difference between a weather forecast and a system that adjusts the climate itself.
Can you share who your clients are?
Not by name: we default to strict NDAs. What we can say: our systems run live in commercial property in Malaysia, alongside systems engineering for a European technology company. The rest we discuss confidentially.
How do you handle data security and privacy?
Our systems run on-premise or inside your private cloud, never on someone else's servers. Operational data does not leave your controlled infrastructure.
What makes QuarkX different?
A rare overlap: Cambridge computational modelling, low-level engineering on global telecommunications infrastructure, and production-scale agent orchestration at a world-leading fintech platform. The person who understands the physics also writes the code that touches your operation.
How do we begin?
Drop us an email. We start with a frank technical conversation about whether this kind of system fits your problem: no commitment, no sales pitch.
What happens if you are unavailable?
Nothing depends on us being reachable. Systems run on your infrastructure, your team can pause or reverse any action without calling us, and documentation is delivered with the work rather than promised after it.
Who owns what you build for us?
Nothing is built without a signed brief: we design against your brief, and the brief stays yours. Ownership and licensing are set out in writing before work starts. Two things never vary: your operational data is yours, and it is exported to open formats you can read without our software.
If the stakes are real, let's talk.
Most engagements start with an exploratory call. No brief needed.