Platform
What you would otherwise have to build.
Four services, one pipeline, and a deployment model that puts all of it in your own cloud account. This page is the technical version of the pitch, because the person evaluating this is usually the person who would have had to write it.
Services
Four services, separately deployable.
Each one owns a single concern and scales on its own. A validation backlog does not stop transmission, and a network outage does not stop intake.
Gateway
The front door. Accepts invoices, resolves the sending participant by TRN, isolates each tenant's data, and owns the status history every trace is read from.
PINT-AE validator
Executes the real UAE Schematron rule set. Returns the specific rule that failed and the field that caused it, not a pass/fail bit.
Peppol node
Transport across the network, participant registration in the SMP, reachability lookups, and message-level status back to the sender.
Application backend
Tenants, users, roles, invoices, trading partners, products, service accounts and webhook delivery. What the two consoles talk to.
The pipeline
Every invoice walks the same eight stages.
These are the actual states the system records, not a simplified diagram. Each one is timestamped and attributed to the service that produced it, which is what makes a support question answerable.
The document lands and is persisted before anything else happens. From here on there is a record even if every later stage fails.
The sender is resolved to a registered participant. An unregistered sender is rejected here rather than three stages later.
Real Schematron execution. A failure is terminal and carries the failing rule ids and the business terms behind them.
Converted to the compliant UBL structure the network expects.
Transmitted to the receiving access point over the network.
The Tax Data Document goes to the Federal Tax Authority. This is the UAE-specific fifth corner.
The receiving access point returns a message-level status: accepted, or rejected with a reason.
Terminal. The document is archived and retrievable for audit.
Validation
Real Schematron, not a shape check.
This is the difference between an access point and a file forwarder, and it is where most home-grown attempts stall.
What it actually runs
- The published PINT-AE rule set, executed as Schematron
- Rule ids and the business terms behind each failure
- Fails closed: a validator that cannot run never returns a pass
- The same rules on inbound documents, driving the status returned to the sender
Why it matters commercially
A rejected invoice is a support ticket. If your platform can only say “invalid”, that ticket becomes a phone call and then an escalation. If it can say which rule failed and which field caused it, the customer fixes it themselves. At a few hundred participants, that distinction is most of your support cost.
Resilience
What happens when something breaks.
Not if. The interesting question for a regulated exchange is what the system does on the bad day.
A worker crashes mid-document
Work is claimed from a durable queue and only acknowledged on success. An interrupted document is redelivered and processed, not lost.
The network is unreachable
Transient transport failures retry on exponential backoff with jitter. Documents queue rather than fail, and drain when the network returns.
A message is delivered twice
Every event carries a deterministic id and every consumer is idempotent, so redelivery is a no-op. This is assumed, not hoped for.
Something fails permanently
It stops immediately instead of retrying forever, lands in a dead-letter path, and shows up as a countable failure in your console.
Scale
Sized honestly.
We would rather tell you what the shape of the system is than quote a number you have no way to verify.
How it scales
Every service is stateless behind its queue, so throughput is a function of how many replicas you run. Validation is the heaviest stage and the one you scale first; transport and reporting scale independently of it.
Storage grows with your archive obligation rather than with throughput, and archived documents move to cold storage without leaving the audit trail.
What we will not claim
You will see platforms advertise throughput figures that no national e-invoicing network in the world produces. Peak load for a UAE access point is driven by month-end and quarter-end batching, not by a sustained rate. The number that matters is whether the system absorbs a large batch without dropping anything and drains it predictably. That is what we will demonstrate, on your data, before you sign anything.
Deployment
Infrastructure as code, into your account.
The entire estate is declared in code: cluster, services, queues, storage, networking and secrets. Deploying to AWS, GCP or Azure is a configuration decision rather than a rewrite, and standing up a second region is a deployment rather than a project.
Your account
Data residency and access control stay with the party that is actually regulated. You.Rebuildable
The environment can be destroyed and recreated from source. That is a routine operation here, not a recovery drill.Yours to keep
Source-code ownership transfer is available where your accreditation or your board requires it.