Victor Luo did the math on his own recruiting once, and it was ugly. Sitting inside TikTok, helping build the recommendation models that decide what a billion people watch next, he and his co-founder had endured something like 800 interviews sourced by outside agencies. The candidates were fine. The process was a machine for wasting engineering time. It was, in a strange way, the exact opposite of everything he did for a living: a system with no memory, no ranking, and no feedback loop.
That contradiction is the origin story of Perfectly, a San Francisco company in Y Combinator's Winter 2026 batch that calls itself an AI-native recruiting operating system. The pitch is simple enough to fit on a sticky note - fill your role in days, not months - but the idea underneath it is the one Luo kept circling: hiring is a recommendation problem, and almost nobody treats it like one.
Figures reported by Perfectly from early customer engagements.
01 / What it isThe recruiter is software, and its name is Paul
Perfectly automates the parts of hiring that usually eat a founder's calendar: sourcing, outreach, screening, and qualification. A hiring team runs an intake session so the system can learn what it actually wants - not the job description, but the taste behind it. From there Perfectly reaches out to thousands of candidates, runs an AI screen that can be reused across future roles, and delivers the ones who make the cut straight into the company's Slack. Behind the workflow sits an AI recruiting agent the company named Paul, which handles the candidate relationship management most agencies do by hand.
The naming is not an accident. Calling the agent Paul turns it into a colleague you delegate to rather than a dashboard you have to babysit - a small design choice that changes how a team relates to the tool.
02 / How it worksFrom intake to interview, in five moves
The screen is the quietly clever bit. Because a candidate's AI screening can be reused for future jobs, the system compounds: every role Perfectly runs makes the next one a little cheaper to fill. That is the recommendation-engine instinct showing through - a pipeline that learns instead of resetting to zero each time.
03 / The problemWhy 800 interviews felt like a bug
Traditional recruiting agencies, in the founders' telling, are slow, expensive, and prone to sending lower-tier candidates for the roles that are hardest to fill. The incentive is misaligned: an agency is paid to produce activity - resumes, intros, motion - not necessarily the one person who says yes. Perfectly's answer is to flip the incentive entirely. The company only gets paid when a role is successfully filled.
04 / The edgeRecommendation systems, pointed at people
What separates Perfectly from the growing crowd of AI sourcing tools is less the interface and more the resume behind it. Luo is joined by Zhuang "Gary" Luo, a former senior machine-learning engineer at Meta with a TikTok stint, and Huimin Xie, who led livestream recommendation for an audience of roughly 100 million users during four and a half years at TikTok. These are people who spent their careers building the ranking systems that match content to humans at scale. Perfectly is that same machinery aimed at the messier problem of matching humans to jobs.
Perfectly says it compresses a typical 20-hour-per-hire recruiting effort down to about one hour.
05 / TractionThe velocity switch
Since joining YC, Perfectly says it has helped fill roles for startups including Giga, Corgi, LlamaIndex, Porter, and Mintlify, along with dozens of others. The most telling data point is a defection. One customer, a Series A founder named Caleb, described dropping two competing tools after a two-week trial.
Speed, in other words, is the feature people quietly switch for. Within its own batch, Perfectly landed a Tier A ranking on the community-run YC Tier List - an early, informal vote of confidence from the crowd that watches these companies most closely.
06 / The model & the marketA five-person recruiting team for dozens of companies
Perfectly sells to fast-growing startups - the ones that need to add engineers faster than they can build an internal recruiting function. It competes against traditional agencies, sourcing marketplaces like Paraform and Juicebox, in-house recruiters, and a wave of other AI hiring startups. Its wedge is the combination the others struggle to hold at once: the judgment of a good recruiter and the speed of software.
There is a second product taking shape on the candidate side. Perfectly has launched Parker, an AI "career super-connector" that lives on iMessage and WhatsApp and connects job seekers with referrals and openings - the beginnings of a two-sided network where the same matching engine serves both the company and the candidate.
Whether Perfectly's numbers hold as it scales is the open question, and the reported multiples are the company's own. But the bet is coherent, and it is the kind only this particular team could make with a straight face: that the least-automated decision at every startup - who to hire - is, underneath, the same problem they already solved for a billion strangers' attention.