Telecom
Radio through core to voice, and the evidence that it works.
What follows is what we have built and can substantiate. Where something is designed but not yet running, it says so.
Where this stands
Delivered, and under proof of concept
The domain proxy and the UE simulator have been delivered to customers and are running in their labs today.
The end-to-end stack — radio through core to voice — has been brought up and run in our own lab with packet captures kept as evidence, and is under proof of concept in a customer lab. It has not yet carried commercial traffic on a live operator network, and we would rather tell you that now than have you find it in week three of an integration.
01
Mobile core and voice
The network behind the radio.
A 4G evolved packet core, a 5G core, and an IMS for voice. Registration has been proven across EPC and IMS against a shared subscriber database, and VoLTE call setup and media captured end to end.
We work at the interface level rather than the box level, which is what matters when a core has to sit beside equipment somebody else built.
02
Radio and protocol engineering
Disaggregated, to the 7.2 split.
A 5G radio access stack covering the central unit — control plane and user plane — and the distributed unit, connected to radios over O-RAN Open Fronthaul 7.2. It has run three-sector fronthaul bidirectionally, with the captures retained.
Below that sits ordinary protocol work: 3GPP layer two and layer three implementation, packet processing, and the timing discipline a fronthaul interface demands.
03
Shared spectrum
A domain proxy between the network and the SAS.
In CBRS a network may only transmit on the channels and power levels a Spectrum Access System has granted it. A domain proxy sits between the two: holding the grants, renewing them before they lapse, and enforcing them on the radios through the management plane.
Ours runs in an automatic mode that mediates continuously, and a manual mode for operators who want an explicit hand on the request. In both, RF legality is enforced by the proxy rather than assumed of the equipment.
There are two implementations — one in Python, one in C and C++ for deployments where footprint and runtime dependencies matter. Both are running in customer labs today.
04
Deployment and operations
Onto the environment the workload needs.
Network functions built to run on bare metal, in containers, or under Kubernetes — chosen against the workload rather than assumed. Real-time paths have requirements that a general-purpose scheduler will not meet, and pretending otherwise is how integrations fail late.
Every component ships with a dimensioning statement: the compute, memory and storage it needs, written into the design rather than discovered during deployment.
05
Verification and assurance
So a claim has a measurement behind it.
A UE simulator built to 3GPP, presenting a population of subscribers to the network across the radio, the core and IMS — signalling and user plane both, at tens of thousands of subscribers. It answers the question a vendor is always asked and can rarely prove: what happens at scale.
Alongside it, functional and interoperability testing, failure injection, soak runs, and lab build-out where a system needs somewhere realistic to run. Results come back as an evidence bundle — logs, captures, configurations, and a manifest mapping each claim to the file that supports it.
The simulator has been delivered to a customer and is running in their lab.
Want to go through it properly?
Architecture documents, dimensioning models and the evidence bundles sit behind an NDA. We are glad to walk through any of it.
Start a technical conversation