Kannan
Kothandaraman

Somewhere around his fifteenth year at Juniper Networks, Kannan Kothandaraman started keeping a mental ledger. Not of wins - there were plenty of those, including leading development of the M120, Juniper's first fully redundant edge routing platform, and eventually rising to Vice President of Product Line Management, overseeing more than $2 billion in annual product revenue. The ledger was of broken things. Specifically, the gap between how network operations were supposed to work and how they actually worked at three in the morning when something went down at a Fortune 500 company.

The pattern repeated everywhere. Too many monitoring tools, each speaking a different language. Data scattered across a dozen systems with no shared context. Alert floods so overwhelming that on-call engineers learned to tune them out. And underneath it all, the fundamental problem: correlation is not root cause. Knowing that three things happened at the same time tells you almost nothing about which one caused the others.

By 2019, after 19 years split between Cisco and Juniper, Kothandaraman had seen every variation of this dysfunction at the world's largest networking companies and their customers. Most executives with that record go consult or join a board. Kothandaraman left to build the fix.

When Kothandaraman and co-founder Nitin Kumar - a 15-year Juniper Fellow who became Selector's CTO - first described Selector's concept in 2019, they called it a "social network of machines and applications." It sounded more metaphor than product spec. What they meant was something precise: a system where infrastructure components could share context with each other the same way people share information in a network, surfacing relationships and dependencies that no single monitoring tool could see.

The timing was deliberately chosen. The technology to actually build it - large language models, scalable knowledge graphs, practical causal inference - was just arriving. Selector was engineered to meet the moment rather than chase it. Today, the platform combines those three elements to ingest data from across an enterprise's network, infrastructure, and applications, detect anomalies, diagnose root causes, and increasingly, resolve issues autonomously before a human ever needs to open a ticket.

The network-specific LLM Selector developed - purpose-trained on network operations data, not a generic model with a prompt layer on top - is a direct product of that philosophy. Generic AI is good at general problems. Network operations has very specific ones, with specialized vocabularies, topology structures, and failure modes that don't appear in the training data for consumer-facing models.

Growth stories usually involve one exceptional year and careful framing of the rest. Selector's is different: four consecutive years of ARR doubling, with 2025 being the most dramatic yet. Cumulative ARR hit 230% of 2024's total. New ARR booked reached 370% of the prior year - meaning the company isn't just growing, it's growing faster, and existing customers are expanding aggressively.

The customer mix is what makes the numbers credible. When 80% of your customers are Fortune 1000 enterprises - organizations with procurement processes, legal review cycles, and vendor fatigue baked in - consecutive doublings mean something different than they do for a product-led self-serve business. Three Fortune 20 companies in manufacturing and healthcare signed in 2025 alone. These are not the organizations that take fliers on unproven technology.

In January 2026, the USPTO granted Selector eight foundational U.S. patents covering the core technical innovations that make the platform work. These are not defensive filings on minor UI improvements. They go directly to the hard problems Kothandaraman identified in his years at Juniper and Cisco.

Kothandaraman's credentials in networking are not the kind you build by attending conferences. He was a software engineer on the M120 - Juniper's first fully redundant edge routing platform - before taking the executive track and spending 13 years as VP of Product Line Management, where he was responsible for Juniper's OS and software, edge and data center routing, and network function virtualization. The teams under him built the infrastructure that Fortune 500 companies actually ran on.

Before Juniper, he did two years at Cisco during the era when Cisco was inventing modern enterprise networking. His academic foundation came from Louisiana State University (BS, Computer Science) and Texas A&M University (MS, Computer Science). The combination of deep systems engineering experience and enterprise product leadership is exactly what makes Selector's pitch credible to CIOs: the founder knows what the ops team is actually dealing with because he spent two decades building the systems they're dealing with it on.

In an industry full of founders promising full automation and the elimination of ops teams, Kothandaraman has been consistent about a different framing. The goal at Selector is "AI that augments, not replaces, engineers." It is a philosophically loaded position, and also a pragmatic one: the enterprises spending real money on operational intelligence are not looking to fire their network teams. They are looking to make those teams faster, less reactive, and less burned out.

The practical version of this is Selector's approach to ChatOps - a conversational interface where engineers can query the platform in natural language, get context-aware responses, and take action without switching tools or parsing dashboards. The next generation of this, per Kothandaraman, involves multi-turn agentic reasoning: an AI that can not just answer a question but pursue a line of investigation, the way a senior engineer would work through an incident postmortem.

His stated mission distills to a single line: "Stop reacting. Start anticipating." The shift from reactive to proactive is the value proposition, the product strategy, and the reason the company exists. Every feature traces back to whether it moves an operations team further from firefighting and closer to foresight.