The Profile
The Operator Who
Counts What Actually Broke
There are 222 trillion combinations of variables that can wreck a single streaming session. Keith Zubchevich knows this number the way a cardiologist knows a patient's resting heart rate - not as trivia, but as the organizing fact of his professional life. Device type, internet connection, app version, backend server configuration, ISP handoff, CDN edge node - mix and match any one of them differently and you get a completely different user experience. Conviva, which he now leads as President and CEO, built its entire existence on measuring that chaos at scale.
Before Zubchevich stepped into the top role in 2021, he spent over a decade as Conviva's Chief Strategy Officer, quietly constructing the infrastructure and partnerships that would make the company the authoritative measurement layer for global streaming. The transition to CEO was less a promotion than a formalization - the company had been running on his strategic logic for years.
But the job he's doing now is different from the one he inherited. Streaming quality-of-experience was the original brief: buffer ratios, startup times, bitrate quality. Clean, measurable, finite. The new brief - analytics for AI agents operating in consumer-facing applications - is messier by an order of magnitude. An AI agent can declare a successful interaction while the customer simultaneously hits the cancel subscription button. Zubchevich sees reconciling that gap as the defining problem of the current technology cycle.
His framework for it is characteristically operator-level. Don't deploy AI agents without first establishing behavioral baselines. Set outcome metrics before launch, not after. Monitor every consumer interaction in real time, not aggregate reports two days later. And never, under any circumstances, accept 80% success on a consumer-facing transaction as an acceptable outcome. "When you launch a consumer-facing agent," he told the Tech Talks Daily podcast in early 2026, "it has to be 100%. Anything short of that is a zero."
The directness is earned. Before Conviva, Zubchevich spent years inside the machinery of enterprise technology at companies that were themselves building new categories. At Cisco, he ran operations for storage-switching and optical solutions, and led the launch of the MDS storage networking technology effort - a product that hit $244 million in revenue within two years. At Riverbed Technology, he was VP of Business Development through the company's IPO, shaping go-to-market strategy and key partner relationships during one of enterprise networking's more consequential growth periods.
Earlier still, he held sales leadership roles at McDATA, Digital Equipment Corporation, and StorageTek - companies that were, at various points, defining what enterprise storage looked like. He absorbed the discipline of selling technology to people who would lose their jobs if it failed. That discipline never left.
Running parallel to his corporate career was a serial entrepreneurial track. Zubchevich founded six companies across telecommunications, content creation and delivery, and internet technology - accumulating more than $500 million in private and public financing and completing over ten acquisitions and mergers across those ventures. The number is striking not for its size but for what it implies about his operating appetite: he has been simultaneously running things and building things for nearly four decades.
He holds a Bachelor's degree in Economics from San Diego State University. It is, perhaps, the most useful undergraduate major for the work he does - Conviva's entire value proposition rests on the idea that the right data, measured correctly, changes economic outcomes for the businesses that use it.
Under his watch, Conviva's sensor network reached 5 billion distributed points collecting data on content, advertising, viewer identification, and engagement - processing 3 trillion real-time events per day. In December 2025, the company launched Digital Product Insights, a platform rewrite for the agentic era that introduced Pattern Analytics (automatic discovery of real consumer behavior patterns), Predictive Intelligence (identifying the strongest outcome drivers across segments, campaigns, and features), Cohort Replay (replacing single-session replays with representative samples from behavioral cohorts), and Nexa (a natural-language interface that lets any team member query the platform in plain English). The Gartner Magic Quadrant for Digital Experience Monitoring recognized Conviva as a Visionary that same year.
The AI agent analogy Zubchevich returns to most often is parenting. He compares consumer-facing AI systems to toddlers handed a job - full of promise, energetic, and completely lacking the contextual judgment to operate safely without guardrails. He uses the Chevrolet chatbot incident, where a dealership's AI was gradually manipulated through iterative requests into agreeing to sell a $76,000 vehicle for one dollar, as the cautionary data point. Not a failure of the AI. A failure of measurement, monitoring, and intervention.
The streaming analogy is also deliberate: "It's going to be messy at first, just like video streaming was in the early days." He watched streaming go from a technical novelty with chronic buffering problems to the dominant global entertainment infrastructure. The early problems weren't solved by better engineers; they were solved by better visibility into what was actually breaking, for which users, under which conditions. That is precisely the instrument he is now building for AI.