Dev Nag

QueryPal started as a complaint. “QueryPal was born out of frustration with how customer support operates,” Dev Nag says. The knowledge a support team needs already exists - buried in documentation, scattered across systems, locked in the heads of senior agents. It just never shows up at the moment a customer is waiting.

So Nag built an AI layer that finds it for them. QueryPal drafts replies from a company's own knowledge base, answers questions through a self-service bot, and shares internal know-how inside Slack - then keeps a knowledge graph that gets sharper every time it is used. The pitch is unfashionably humble for an AI company: he is not here to fire the support team. He is here to stop them doing the boring part.

The market he is aiming at is enormous. Nag pegs global customer-support operations at a $300-400 billion opportunity, and QueryPal's public claim is blunt: cut up to 60% of support cost. The proof point he likes best is smaller and more human - an early pilot that cut a customer's average response time in half and won an 85% approval rating from that customer's own agents. When the people whose jobs you are automating give you 85%, you are doing something right.

The product line reads like a single idea pointed in four directions. QueryPal Resolve drafts email responses straight from the knowledge base. QueryPal Guide is a web-based Q&A bot for customers who would rather help themselves. QueryPal Chat lives inside Slack so an agent never has to leave the conversation to find an answer. And the QueryPal API drops that same question-answering into a company's own systems. Different surfaces, one promise: the right answer, already written, at the moment it is needed.

None of that ships without trust, and enterprises do not extend it cheaply. So a chunk of the early work was unglamorous and load-bearing - SOC2 compliance, integrations that read a company's systems without rewiring them, and a careful answer to the fear every support leader voices first: is this thing here to replace my team? Nag's response is consistent to the point of stubbornness. The agent stays. The drudgery goes.

Long before “agentic AI” was a slide in every deck, Nag was wiring software into live systems where mistakes are expensive. He was an engineer at eBay, PayPal, and Google, and the founding engineer at GLMX, the electronic securities-trading platform for money markets - the kind of place where a millisecond and a misplaced decimal both matter.

Then he built his own. Wavefront brought real-time monitoring and analytics to cloud infrastructure, the dashboards and alerts that tell engineers when something is on fire and where. He ran it as founder and CTO until VMware acquired it in 2017. Inside VMware's Office of the CTO he didn't coast - he launched the company's flagship AIOps product, pointing machine learning at the operational chaos he had spent a career mapping.

His next act, CtrlStack, emerged from stealth in December 2022 with a $5.2 million seed round co-led by Sequoia Capital and Engineering Capital. The product unified change events from AWS, Kubernetes, PagerDuty and more into a single “Event Timeline,” so teams could see what changed right before everything broke. Sequoia's Bill Coughran, who backed him across companies, summed up the through-line in five words: closing the gap between cause and effect.

That phrase is not marketing varnish - it is genuinely the question Nag keeps circling. He once wrote an essay titled “Root Cause Analysis: How can an idea that's wrong be so useful?” - a working engineer's meditation on the comforting fiction that every outage has a single, findable cause. Real systems fail for tangled reasons. The honest job is not to crown one villain but to narrow the search until a human can act. CtrlStack was that belief turned into software, and it is not hard to see the same belief humming underneath QueryPal: surface the relevant thing fast, let a person make the call.

Ask the internet about Dev Nag and a surprising amount of it points not to a funding announcement but to a tutorial. “Generative Adversarial Networks (GANs) in 50 lines of code (PyTorch)” - a Medium post paired with a tiny open-source repo - became one of the friendliest on-ramps to a famously slippery idea. Two neural networks battling over a shifted Gaussian, explained so a beginner could actually follow along.

It tells you something about how he thinks. The math-and-psychology Stanford double major has published papers in mathematical biology and medical informatics and holds more than a dozen patents in AI and security - and still his instinct is to strip a hard thing down until a stranger can rebuild it. He mentors startups across AI, technology, and biotech for the same reason. The teacher and the founder are the same person.

The two Stanford degrees turn out to be less of a contradiction than they look. Math gives him the machinery; psychology gives him the customer. A support ticket is a math problem - retrieval, ranking, the right document at the right moment - wrapped in a human one - a frustrated person who wants to feel heard. Most AI founders are fluent in only the first language. Nag studied both on purpose, and it shows in how he talks about the work: less about model benchmarks, more about how a tired agent feels at 4pm on a Friday with forty tickets in the queue.

The endgame Nag describes for QueryPal is less about software and more about what software frees up. Automate the repetitive research and drafting - instant ticket resolution, API integrations, a knowledge graph that keeps learning - and the support agent stops being a human FAQ. “We aim to elevate agents from reactive responders to strategic contributors,” he says, the ones building real relationships and holding onto the accounts that matter.

It is the same move he has made his whole career, just pointed at a new department. Wavefront freed engineers from staring at dashboards. AIOps freed operators from chasing alerts. QueryPal wants to free support agents from the inbox. The machine takes the grind; the human keeps the judgment.

His theory of leadership has bent to match. He used to think running a company was about being the technical authority. Building Wavefront taught him otherwise. “It's actually about understanding what motivates each person on your team and connecting their individual aspirations to something larger,” he says. These days he describes most of his job as translation - carrying customer pain to the engineers, technical possibility to the customers, and market reality to the investors. For a founder who came up writing fraud systems and monitoring engines, it is a quietly radical admission: the hardest system to debug is the one made of people.

Step back far enough and Dev Nag's resume stops looking like a list of jobs and starts looking like one continuous argument. Fraud detection at PayPal, search and scale at Google, a trading platform where latency is money, a monitoring company, an AIOps product, an observability startup, and now an AI support agent. Different industries, same shape: take a flood of signal, find the part that matters, and hand a human a decision they can actually make.

It is an unfashionable kind of consistency in a market that rewards pivots and reinvention. Nag keeps showing up to the same problem with sharper tools. The patents pile up, the companies change names, the backers re-up - and the through-line holds. He is still, two decades in, closing the gap between cause and effect, and still convinced the point of the machine is to make the person better at being a person.