Built by Telperion
Proof, not a product catalog.
We don't sell a platform — every engagement is built for the client's network. These systems were built for and run in real ISP operations, and they show what that bespoke work looks like.
Netmon monitoring suite
Fragmented monitoring replaced with one operator view across a Juniper-heavy service-provider core.
Multi-source by design
NETCONF, SNMP, JFlow/IPFIX, syslog, and optical telemetry correlated on a live topology, so operators can tell access from core from transit from optical.
A suite, not a monolith
Separate modules for core, customer edge, flow, customer support, and software/version tracking with CVE checks, sharing single sign-on and a NetBox source of truth.
What it shows
Telemetry pipelines, data modelling, and operator UX for networks where downtime has a cost — the groundwork every useful AI agent needs.
AI-assisted jump host
A browser replacement for PuTTY with an assistant that reads the session and proposes commands.
The human is unconstrained; the AI is constrained
Engineers keep full interactive access under their own device credentials. The assistant is fenced in by a per-vendor, per-tier policy gate and an approval step.
Commit is never the AI's to make
The assistant may stage configuration where a device has a candidate. Committing to the running device is never something it can propose — in any spelling.
Attributable by default
Users authenticate to devices as themselves, never through a shared account, so every action lands in AAA logs under a real name.
Event correlation & AI alert triage
One search surface over syslog, flows, and DHCP, and a home for AI alert decisions that previously had none.
Storm to incident
An incremental correlator groups related log lines into events and keeps a reference to every source line as evidence.
AI decisions on the record
Each triage records the input, model, verdict, confidence, evidence, and the operator's final disposition — so the AI can be audited and measured.
Layers kept separate
Parsing, correlation, the agent, and the operator view are distinct layers, so each can be replaced without dragging the others along.
Want something like this for your network?
Start with a fixed-scope assessment of your tools, scripts, and alert noise.
About the Assessment