Margareth Felicia Gono, 24, started last month as a senior associate at DoorDash, which is the happy ending. The interesting part is how she got there, which is that she treated the worst stretch of her professional life as a dataset. As she told Business Insider, she grew up in Indonesia, finished a master’s degree in applied analytics in May, and needed a visa to stay in the US — her parents had paid for a private university and then an Ivy League master’s on top of it. “I felt I had to apply to anything that might relate to my experience, even if it wasn’t the right fit,” she says. It was, in her word, scattershot.
So she started recording what happened. What roles got responses? What patterns showed up? Her rationale deserves to be quoted because it is the whole philosophy in five words: “Anxiety is unprocessed data. I was tired of feeling bad about my search and figured I might as well learn something instead.”
The instrument she used was Anthropic’s Claude, deployed “almost like a partner.” She would paste in a job description and ask Claude to identify the main skills, keywords and themes — then paste in her résumé alongside it and ask, “What top keywords am I missing?” She describes it as her own version of an ATS score, which is a pleasing bit of symmetry: employers use software to grade candidates, so she built software to grade herself first. She is careful about the limits. “I didn’t want AI to invent experience or make me sound like everyone else. I wanted it to show me where my experience wasn’t being communicated clearly enough. Sometimes that meant changing one bullet. Sometimes it meant realizing that applying would be a waste of time.”
The funnel
She also used Claude as a career coach and concluded she likes three things — talking to people, making things pretty, and finding patterns — which became three tailored résumés: one for analytics, one for sales and marketing, one for consulting. Eventually she turned the whole exercise into a dashboard, and here it is, because the numbers are the story: 186 applications, 44 cold outreaches, 19 interviews, six take-home assignments, three final rounds, one offer. That is an interview rate of about 10%, and a conversion rate on the whole process of roughly half a percent — which, reframed, means 185 rejections. “Seeing it that way made the process feel less like 185 personal failures, and more like a funnel I could keep improving,” she says.
The interviews themselves got the same treatment. When recruiters mentioned they were using an AI note-taker, she would ask if she could record too; most said no, which is a fun asymmetry — the company’s robot may take notes on you, but your robot may not. From the yeses, she would scrape the notes and run sentiment analysis: where she hesitated or used filler words, where the interviewer pushed for specifics, where they seemed engaged, agreed or laughed. When recording was refused, she wrote down everything she remembered immediately afterward.
She also built a Claude skill to scrape the LinkedIn profiles of her interviewers — she copies everything on the page, pastes it in, and since “Claude knows who I am because I talk to it a lot,” asks it to find similarities. In one case she connected with a recruiter over a cheese club they had both belonged to, years apart. “That got me further than any cover letter or résumé font choice ever did,” she says — which is, of course, the oldest finding in job-hunting, rediscovered by the newest tool: the personal connection is the whole ballgame. She also swears by the questions at the end of an interview, asking how people got where they are and what advice they’d give a new graduate. “It’s easy to have imposter syndrome,” she says, “but ... if they don’t see my value, someone else will.”
Her own conclusion is the one she has earned the right to make: “Everyone has access to AI now, so that isn’t a differentiator. What makes the difference is knowing what to ask, how to interpret what AI gives you, and where judgment still matters. AI can help you see patterns, but you have to decide what they mean. It can help you get your foot in the door, but it can’t make you want the job or make someone remember you.”
And there is one number her dashboard did not capture, because it has no column for it: her offer, number 187 in the counting, is the only one of the 186 applications that mattered. The machine can count the funnel. It cannot pick the winner.