Selector gains eight US patents for AI-driven observability

Selector has received eight US patents covering techniques in causal inference, large language model training, AI-based correlation, predictive maintenance and network path intelligence for network operations and observability.

The company said the patents span methods for identifying root causes of incidents across network domains, structuring signals from system logs, forecasting capacity and tracing packet paths at a specific time. Other patents cover storage and querying of network parameters, reporting that accounts for maintenance windows, and early identification of optical transceiver failures.

Selector operates in the AIOps and observability market, where suppliers focus on software that analyses telemetry from networks and infrastructure. Network operators and large enterprises often use these tools to track incidents, investigate performance problems and manage capacity across hybrid environments.

The patents include a method described as "Root Causation for Network Operations". Selector said the approach applies causal reasoning to identify the origin of faults in multi-domain environments. The company also listed a patent focused on "Metrics, Events, Alert Extractions from System Logs". It said the method converts unstructured telemetry into structured information for analysis.

LLM training

One patent covers a method described as "Dashboard Metadata as Training Data for Natural Language Querying". Selector said it uses metadata from visualisations and user interactions as training data for a network-focused large language model. The company said this supports natural language querying in network operations workflows.

The patent list also includes "Maintenance Window Aware Reporting". Selector said the approach detects and excludes maintenance windows from performance analytics. It said this changes how service availability and reliability metrics are calculated and reported.

Network paths

Several of the patents focus on network topology, routing state and path analysis. Selector listed "Methods and Apparatus for Efficient Storage and Querying of Communication Network Parameters". It said the method addresses storage, indexing and querying for large-scale network information.

Another patent, "Methods and Apparatus for Determining a Path that a Data Packet Would Traverse Through a Communication Network at a Time of Interest", covers reconstruction of packet paths at a specific point in time. Selector said the approach supports historical analysis and forensic investigation. It also said it improves the accuracy of root cause analysis based on the reconstructed path.

The company also cited "Methods and Apparatus for Network Tracing, Forecasting, and Capacity Planning". It said the method uses analytics and modelling to predict capacity risks and network growth trends. The patent list includes "Early Identification of Optical Transceiver Failures", which Selector said uses predictive modelling to detect hardware degradation.

Company comments

Selector said the patents reflect work on applying machine intelligence to network data, events and dependencies. It said the inventions relate to parts of its platform including correlation, natural language interaction and analytics.

"These patents reflect years of focused innovation to bring AI and causal reasoning to the heart of network operations," said Nitin Kumar, CTO and Co-founder, Selector. "Selector's platform doesn't just monitor data, but actually understands relationships, predicts failures, and explains why events occur. These innovations are foundational to how we're reimagining observability for the AI era."

The company said the patents relate to a framework that links cause and effect across distributed systems. It said the work sits alongside its use of a network-trained language model and correlation methods.

"Selector's patent portfolio represents a step forward in how AI reasons about network data," said Surya Nimmagadda, Chief Data Scientist at Selector. "Our goal has been to move from statistical correlation to genuine causal understanding-teaching machines to think like engineers. This body of work is the result of rigorous experimentation in applied AI, graph analytics, and knowledge representation."

Selector said its customers include telecommunications providers, cloud service providers and large enterprises. The company said it has investors including Two Bear Capital, Atlantic Bridge Ventures, Sinewave Ventures, Ansa Capital, Singtel Innov8, Hyperlink Ventures, AT&T Ventures, Bell Ventures and Comcast Ventures.

Selector said it will continue work on causal reasoning, natural language interaction and predictive analytics for network operations teams.