# Nobody's Done This Job Before. That's the Job. **Author:** WEKA **Published:** August 26, 2026 ![Black and white image of a man looking out an office window at a city skyline, with blurred coworkers at desks in the background.](https://cdn.sanity.io/images/ult5g8gw/production/876a01916072cd3f2b638ae4bf35903b84835de2-1408x768.jpg) ## TL;DR Explore how AI hypergrowth is reshaping recruiting, culture, and people leadership in this episode of Deep Geeks. - Industry leaders are hiring around AI fluency, potential, and culture fit over pedigree. - Recruiting’s repetitive first mile is getting automated to focus on interviews and fit. - Engineer culture as core infrastructure driving real AI outcomes at scale. Capital can show up in a wire transfer. Culture can't. That's the problem sitting underneath every AI infrastructure company right now. The org chart is scaling faster than anyone can define what belongs on it. There's no résumé filter for "has built a people function through 1,000%+ headcount growth in under two years." That job doesn't have a job description. It doesn't have a playbook. Which is exactly why the newest episode of Deep Geeks is worth your next 35 minutes. Dr. Serena Huang sits down with Roopesh Panchasra, VP of Talent Acquisition at Nscale, and Iris Hermelin, Chief People Officer at WEKA — two people leaders who aren't theorizing about hypergrowth from a conference stage. They're living it, at companies where headcount changes meaningfully between the Monday standup and the Friday all-hands. ## **Where hiring decisions get made** The hiring problem isn't that “qualified candidates” are scarce. It's that AI has rewritten the recruiter's job description. Roopesh doesn't hedge on this: > "It's completely transformed the way we go about our end-to-end recruiting process, and this is just the very beginning. Competitive market intelligence, candidate sourcing, screening at volume — we've had to automate all of it. That repetitive first mile. Getting that out of the way gets us to where hiring decisions actually have to be made: the interview, the feel, the fit, the culture. I'd go as far as to say the role of the recruiter in its current form will never be the same. And this is still very early days." The data backs him up on the "hire for potential over pedigree" instinct too. This isn't just a WEKA and Nscale philosophy. [Microsoft's Work Trend Index](https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part) found 71% of leaders now say they'd rather hire a less experienced, AI-fluent candidate than a more experienced one without those skills, and more than three-quarters expect early-career talent to be handed real responsibility sooner. The "safe" hire, the one who's done the exact job before, at a company you admire, might be the riskier bet in 2026. On the flip side, Serena raised the point that candidates are using AI too, and that's created its own arms race. > "We have a tool that helps us identify applications that are unreal, through various data points," Roopesh says. "We had one applicant apply 100 times for different positions. Same person, same contact details, different personas each time." Iris adds that at WEKA, ATS systems now flag likely-fake profiles automatically, which is part of why both companies lean so heavily on employee referrals and direct sourcing rather than open applications. ** ** ## The multiplier nobody puts on a dashboard If there's one idea from the conversation that deserves to outlive the episode, it's this: culture isn't a values poster. It's infrastructure. And at scale, it has to be engineered as deliberately as anything else in the stack. That's not just a nice sentiment either. Microsoft's own research found that organizational factors like culture, manager support, and talent practices account for more than twice the measurable impact on real AI outcomes as individual skill or effort. Roopesh and Iris are describing, in plain language, exactly the variable the data says matters most. ## Why watch it AI infrastructure companies love talking about GPUs, power, and supply chains as the ceiling on how fast they can grow. This conversation makes the case for a ceiling in a different room of the house: whether you can hire people who've never done the job, and whether the culture that got you to 100 people can survive becoming, in Iris's words, essentially a new company every six months. Nobody's written the job description for this role yet. Roopesh and Iris are writing it live, building roles for the builders of AI. Watch [the full episode here](/podcast/nobody-has-done-this-job) or find it on [Spotify](https://open.spotify.com/episode/6p1K8NVCezRmSgsqg22Sd0?si=5e35dc77ccce4b9f) and [Apple](https://podcasts.apple.com/us/podcast/nobodys-done-this-job-before/id1896864991?i=1000785975749) Podcasts — make sure to Subscribe to Deep Geeks to keep up with the emerging trends in AI Infrastructure.