Victor Luo Wants Hiring to Work Like a Recommendation Engine

The first thing to understand about Victor Luo is that he spent the formative years of his career watching a machine get very good at guessing. At TikTok, he worked on the recommendation systems that decide, in a fraction of a second, what shows up on your screen next. It is a strange kind of magic - the software has never met you, and yet it seems to know your taste better than your friends do.

Then he would close his laptop and walk into a hiring meeting, where the most important matching problem a company ever faces - which person belongs in which role - was still being solved with cold emails, keyword searches, and a lot of guessing that was not very good at all.

That contradiction is the whole story. Luo is the CEO and co-founder of Perfectly, a company in Y Combinator's Winter 2026 batch that describes itself as the first AI-native recruiting operating system. His pitch fits in a single sentence, and he repeats it often.

01 - The 800 interviewsThe number that would not leave him alone

Every founder has an origin number. For Luo and his co-founders, it was 800 - the count of agency-sourced interviews they collectively sat through during their time at TikTok. It was not a happy number.

"After working at TikTok on the world's most advanced AI recommendation, we couldn't believe how outdated our recruiting process was in comparison," Luo has said. The pipelines moved slowly. The candidates were rarely top tier. And when a role got genuinely hard to fill, he found that most recruiters "would rather ghost me" than do the work.

You can hear the engineer's specific frustration in that. Here was a person who spent his days building systems that compound information - every click, every watch, every skip making the next prediction sharper - forced to hire through a process that seemed to lose information at every step. A resume gets summarized. A summary gets skimmed. Context evaporates. By the time a candidate reaches a hiring manager, most of what made them interesting has been sanded off.

02 - The buildAn OS, not another agency

Luo did not set out to start one more recruiting shop. Perfectly is pitched as an operating system that automates the full function - sourcing, outreach, screening, and qualification - so that interview-ready candidates simply show up in a hiring manager's Slack. The team's stated obsession is narrow and stubborn: the highest quality candidates, and nothing that gets in the way of that.

The centerpiece is an AI recruiter the company named Paul. The idea is that Paul carries what Luo calls infinite memory and context - it does deep research into each candidate, and rather than stopping at surface-level experience, it tries to predict interests, trajectory, and how someone will actually perform in an interview and on the job. "Paul doesn't need a mountain of historical data to understand what you're looking for," Luo says of its cold-start ability.

The reported results are the reason people started paying attention. Perfectly points to hiring that runs roughly four times faster, candidate volume around ten times higher, and interview pass rates about double what a traditional agency delivers. Behind the scenes the team cites even sharper numbers - recruiting effort dropping from twenty hours to one.

"By removing the human bottleneck, we provide candidates at up to 10x the volume," Luo says. "The amazing thing is, our candidates are 2x more likely to pass interviews." And Perfectly only gets paid when a role is actually filled - a success-based model that puts the company's confidence where its mouth is.

That client, according to the company, cancelled its Paraform and Juicebox subscriptions and walked away from every other agency, citing the sheer velocity. Early logos include Giga, Corgi, LlamaIndex, and Porter.

03 - Paul and ParkerTwo agents, two sides of the table

Perfectly started on the hiring-manager side of the market, with Paul. But Luo and his co-founders kept running into a pattern that bothered them for a different reason - not slow companies, but stuck people. Brilliant candidates sitting in the wrong roles simply because no one had connected them to the right one.

So they built a second agent, this one for candidates, and named it Parker. Parker lives on iMessage and WhatsApp and behaves less like a job board and more like a well-connected friend who happens to know everyone in tech. It maps out who you should be talking to, drafts outreach that actually sounds like you, and - because companies come to Perfectly to hire directly - lets opportunities start finding you instead of the other way around.

There is a stat Luo likes to cite here: something like 70 to 80 percent of tech jobs get filled through referrals and warm intros before they ever hit a job board. If you are not in the right room, he argues, your value goes unnoticed no matter how good you are. Parker's job is to get you into the room. "Text Parker," the launch note reads. "Don't be surprised when it knows things about yourself that you don't."

04 - The throughlineFrom a research lab to a Tier A ranking

Luo studied computer science at the University of Virginia, where his path ran through the school's engineering research world and its Link Lab. Before Perfectly he moved through Amazon and an AI venture called Synthminds.AI, and then TikTok, where the recommendation work would later become the intellectual seed of everything he is building now. His public founder bio is characteristically blunt about the mission: "I help startups hire 11x engineers."

He is not doing it alone. Perfectly's founding trio all overlapped at TikTok - Zhuang, who goes by Gary, Luo, a former machine learning engineer at Meta and TikTok, and Huimin Xie, a former tech lead and senior ML engineer whose own bio reads "building the future of human careers." Three people who spent years inside one of the most sophisticated matching machines ever built, now aiming that instinct at the labor market.

By mid-2026, the bet was paying off in the way early bets do - not in certainty, but in signal. Perfectly hit roughly a million dollars a year in traction during its YC batch, one of a small handful of teams to reach the milestone, and earned a Tier A ranking. For a company still measured in months, that is less a finish line than a green light.

05 - The wagerWhat he is really betting on

Strip away the agents and the metrics and Luo's thesis is almost philosophical. The best human recruiters, he would argue, win on two things: memory and taste. They remember every candidate, every conversation, every subtle signal of fit. And they have a feel for who will thrive where. Software has always had a clear edge on the first - infinite memory is cheap. Luo's real wager is that the second thing, taste, can be learned the same way a feed learns yours: from feedback, at scale, getting sharper with every loop.

If he is right, recruiting stops being a numbers game of cold outreach and becomes a prediction problem - the same category of problem he spent years solving for content. If he is wrong, well, he will have the data to know it quickly, which is rather the point of building it this way.

For now, the number that matters most to Luo is still that original one. Eight hundred interviews taught him what a broken system feels like from the inside. Everything he is building is an argument that it does not have to feel that way - for the companies doing the hiring, or for the people quietly hoping someone finally notices what they are worth.

WatchPerfectly in Luo's own words