Built on real science.
Protected by real patents.
Reach's technology isn't borrowed from the standard telco playbook. It's invented. Our IP portfolio begins with a granted patent covering a core innovation in mobile network intelligence: crowdsourced load sensing, neural-network-driven prediction, and intelligent data scheduling.
Crowdsourced mobile network load measurement and predictive scheduling
View on Google PatentsA smarter way to understand network load, built from the ground up
This patent describes a system for measuring and managing mobile network load by crowdsourcing data from client devices rather than relying solely on network-side infrastructure. A central cloud-based control server coordinates the process by sending each device a probing profile: instructions specifying what to measure, when, and how often.
Each client collects radio access network (RAN) metrics, app performance data, and packet-level measurements including delay, jitter, and loss. That data flows back to the server as uplink probes. The server processes all incoming data through a pre-trained neural network to estimate current and near-future network load, identifying peak congestion periods and low-usage windows.
Based on that analysis, the server sends scheduling information back to devices. Apps can defer large data transfers to low-congestion windows, reducing interference and improving performance for everyone on the network. No infrastructure upgrade required on the carrier side.
Three layers.
One patented system.
The patent covers the full stack: sensing, inference, and scheduling. Each layer depends on the others.
Devices as network sensors
Each phone or tablet runs a probing profile issued by the central server. The profile specifies exactly what to measure, when, and how often. Devices collect RAN metrics, app performance data, and packet-level measurements including delay, jitter, and loss, then send it all back as uplink probes. The network gets a live picture of conditions from the inside out.
Neural network load inference
All incoming probe data feeds into a pre-trained neural network on the cloud server. The model estimates current network load and predicts near-future peak and valley windows with a level of precision that network-side infrastructure alone cannot achieve. The result: the server knows what's coming before it arrives.
Intelligent data scheduling
Once the server maps load patterns, it sends scheduling information back to devices. Apps defer large transfers like video downloads to low-congestion valley windows. Less interference for active users, better throughput for deferred tasks, and no changes required to the carrier network. The system coordinates at the edge, not the core.
The more congested the network,
the smarter the system gets.
Most network sensing approaches degrade under load. This one doesn't. As more users fill a network sector, more devices contribute probe data to the cloud server. The neural network gets more signal precisely when accurate readings matter most. The system scales with the problem it's solving. That feedback loop is the invention.
Ready to build on
a real platform?
The technology is solid. The team is here. Let's talk about what launching a mobile service under your brand actually looks like.