However, his drive to eliminate communication friction and ensure people are “truly understanding each other” is rooted in a personal journey that stretches back before his time at Uber and Duolingo—to when he learned how barriers can keep people from communicating.

Born in the UK to parents who had immigrated from India, Verma spent his young childhood in the midlands city of Sheffield. He recalled in a recent interview with Fortune that, for the longest time, he was “the only Indian kid in my class.”

While his hometown is “a great place,” he couldn’t help but wonder “what the rest of the world felt and looked like.” He recalled how he was fascinated with a physical atlas in the days before Google Maps. “I’m like just looking at maps and drawing places based on that.”

Years later, when Verma was an engineering major at Carnegie Mellon University in Pittsburgh, he heard Google’s head of engineering, Alan Eustace, give a talk about ongoing projects, including Google Translate. “I was like, ‘Wow, this is the future of connecting people.” Within a few years, Verma was working on Google Translate, part of a journey that took him to Duolingo as its first ever head of product, before a stop at Uber, also as head of project, and a return to Google.

Verma told Fortune that he loves his work because he has always loved the idea of something having to do with travel, saying he wanted to be a travel show host when he was a kid.

“I really liked the appeal of meeting people in different cultures and then truly understanding each other,” he said, adding that he’s been lucky to do a job that lets him do the same thing.

Beta-testing while driving the machine

As head of Google Meet, Verma told Fortune that he is always beta-testing his product, which billions of users depend on to do business. (Although more specifically, he said he’s always testing the product internally, from early research proof-of-concepts to end-user prototypes and beta.) He said the beauty of Google Meet is that it’s part of Google Workspace, which means that all the notes you take on it become part of your Google Drive.

Verma said he uses Take Notes with Gemini in “pretty much all of my meetings.” With a single click, this tool instantly creates a live Google document of meeting notes, which become the team’s “decision of record,” transforming discussions into something “durable and sticky.”

He noted that this reliability has made distributed collaboration easier, recalling a time when a third of his team in Stockholm felt confident skipping an inconvenient meeting, a huge change from just a few years ago. The team was trying to get some feedback from some Bay Area-based executives, but the timing just wasn’t right. Nevertheless, they felt confident in the meeting going on without them. “We trust that you’ll represent our point of view well and we’ll read the notes and we’ll read the transcript later on.”

The Google Meet chief also relayed that he often debriefs on projects, and he asks team members what went well and what could have gone differently. Once all of those are written on a whiteboard, he cross-references them with what the Gemini AI notetaker thought of the meeting. “When in doubt, you can go back and read the transcript,” he said, noting that Gemini includes citations, so the minute anything is in doubt, “there’s a one click to that part of the transcript where you can then just read what happened there.”

The goal is to use AI to facilitate the high fidelity of human conversation, including tone and emotion, allowing teams to reach resolutions faster than asynchronous communication. Verma illustrated this by recounting a 60-minute discussion with an engineering counterpart. By using Ask Gemini to summarize, they immediately produced a 15-bullet product specification.

Real-time translate

The innovations from Verma and his team spring to life when Fortune talks to Niklas Blum, a native German speaker based in Google’s aforementioned Stockholm office. He demonstrates Google’s real-time speech translator, a very different product from Google Translate.

“Don’t you hear me, Nick? I’m speaking German to you now,” Blum said, as this journalist could hear him speaking in the background in something that sounded Germanic, while, with a few seconds delay, his voice came through the speakers in English, somewhat uncannily.

Blum explained that AI technology has now progressed to the point where it clones his voice in real time to make it sound like he is talking in English.

He added that Verma’s team worked closely with Google’s DeepMind on the various technologies, with multiple layers of AI working: the translation plus the generation of the translated voice. To some extent, he said, the lag is very dependent on the language being translated, since the AI has to account for the grammar and complexity of the language being used. German often puts the verb toward the end of the sentence, for instance, so there is a lag as the AI makes sure it has the correct meaning.

This real-time translation tool came together over the course of about two years, Blum said, adding that it began as an exploration without a certain deadline. But the more they communicated with global companies that do business across language barriers, they saw a need in the marketplace.

He said it was “really hard” for team conversations across different languages to be held while making sure that everyone received all the correct information. “We want Google Meet to not only be a tool to connect people, but [something] that creates value in the conversation,” he added.

“The beauty of technology is not just the fact that it can automate and make it easier,” Verma said, but that you can trust “that it truly is neutrally representing everything that was said.”

There’s an imperative behind this, of course, somehow providing a fix to the post-pandemic onslaught of endless meetings that define most workdays.

Verma notes that these are often ineffective, leaving participants unsure of what was decided and what comes next. He said he feels a real need to address “meeting fatigue” that comes from meetings just not being run well.

Verma affirms that every improvement is filtered through the question, “how could we have helped them do this better?”