One place for every job
Bookings from every channel land in the same queue. Nothing depends on a notebook or on who took the call.
Work
Things we built
that are still running
Client work is described by what it does and what it changed. Clients are named where they agreed to it — but the technical detail of a network is never published, for any of them.
Case study · Field service
AirconLab — booking
and dispatch platform
An aircon service business taking jobs by phone and text, tracked on paper. Every booking depended on whoever answered remembering to write it down and pass it on.
The problem
Bookings arrived through three channels and lived in a notebook. Double-booked technicians, jobs that were never assigned, and no way to answer "when is someone coming?" without calling the office.
Growth made it worse, not better — more jobs meant more chances for one to fall through.
What we built
- Booking
- Customers request a service online, choosing the type of job and a preferred slot
- Dispatch
- Jobs assigned to technicians against real availability, so double-booking is not possible
- Job record
- Each visit carries its own history — what was serviced, by whom, and when
- Follow-up
- Service history per unit, so repeat maintenance is scheduled rather than remembered
Technicians are not double-booked
Assignment is checked against availability at the moment it is made, not discovered on the day.
Service history that survives staff
What was done to a unit last visit is on the record, not in the memory of the technician who left.
Network builds
Four networks in operation
Named where the client agreed to it. What we never publish — for any of them — is topology, addressing or equipment inventory. Those are the details that would tell someone where to start, and no portfolio is worth handing them over.
Hap Suy Hardware Co.
Trading & distribution · 4 branches
Four-branch wide-area network
Four sites that had been operating as four separate islands, each dependent on a single internet service. We gave every branch a second line with automatic failover and joined all of them with encrypted site-to-site links, so warehouse and ERP traffic reaches the central systems continuously rather than when the connection happens to be up.
The same links carry the support path, so faults are diagnosed and fixed from the office instead of by sending someone across the city.
Branches keep trading through ISP outages
All sites reach central systems
Remote support replaced site visits
Camera systems reachable from head office
Lazada
E-commerce & logistics
Fulfilment site connectivity
Network engineering for a fulfilment operation where an hour offline is measured in undelivered orders — redundant paths for order traffic, with the operational network kept separate from office use so one cannot degrade the other.
Loss of one path is not an outage
Operations separated from office traffic
Commercial property
& retail group
Multi-tenant estate network
Tenant, operations and management traffic separated on one estate, with a redundant core and access between departments granted by rule rather than by default.
Tenant traffic isolated from operations
No single link failure takes the estate down
Hotel group
Guest and back-office separation
Guest wireless genuinely separated from property operations, with an administrative path that never crosses into guest space.
Guest devices cannot reach operations
Remote administration without site visits
Infrastructure in operation
The platform everything
else sits on
Networks carry the traffic, but the servers, directory and storage are what actually hold the business. This is the layer we design, run and recover — day to day, not just at installation.
Directory services
Windows Server · Active Directory
Domain, identity and access
Active Directory as the single source of truth for who someone is and what they may reach — domain controllers with DNS and DHCP, Group Policy applied per department, and RADIUS so the same account controls VPN and wireless access.
Including domain controller migration to new hardware without an interruption users could notice, and disabling a leaver in one place instead of eleven.
One account, one place to revoke it
VPN and wifi authenticated against the domain
Policy applied consistently, not per machine
Controllers migrated with no user-visible downtime
Virtualisation
Proxmox VE · VMware vSphere
Clusters, replication and failover
Production consolidated onto clustered hypervisors with shared quorum, storage-level replication to a second machine on a short interval, and high availability configured so a failed node restarts its workloads rather than waiting for someone to notice.
Also the unglamorous half: right-sizing over-provisioned machines, snapshotting before every change, and physical-to-virtual conversion of ageing servers that could no longer be trusted on their original hardware.
Standby copy kept minutes behind production
Automatic restart of workloads on node failure
Ageing physical servers converted to VMs
Snapshot before change, rollback in seconds
Storage & backup
NAS · ZFS
Where the copies actually live
Network-attached storage and ZFS pools sized for the data rather than the invoice, with scheduled backups, retention that does not quietly fill the disk, and an off-box copy so one dead machine cannot take both the original and the backup.
Backups are opened and restored on a schedule. That is the entire difference between a backup and a hope.
Off-box copy on separate hardware
Retention that will not fill the array
Restores tested against the real archive
Capacity monitored before it becomes urgent
Automation in production
Running quietly in the background
None of these need anyone to remember them, and all of them report that they ran.
Daily infrastructure report
Every morning, a health summary of every server, container and link arrives in the team chat — read before anyone opens a console, so problems are found before users report them.
Meeting scheduler
Meetings booked from a chat message: the bot creates the conference, puts it on the calendar and sends the link to everyone who needs it.
Request and alert routing
Approvals and alerts delivered to whoever is actually on duty, instead of an email chain that stalls when one person is on leave.
Backup with verified restores
Scheduled backups whose archives are opened and tested on a schedule, plus an off-box copy — because a backup stored beside the thing it protects is not a backup.
Replication to a standby
Production replicated to a second machine on a short interval, with automatic failover, so a hardware failure is an interruption rather than a rebuild.
Monitoring and config backup
Service health checks with escalation, and every device configuration backed up and versioned so a change can always be traced and reversed.
References available on request
Want to talk to someone
we already work for?
We will arrange an introduction, with their agreement.