Handshake Activates PhD Talent Network
Handshake vs Mercor
Handshake’s edge is not just that it found a hot AI market, it is that it already owned a verified supply of exactly the people frontier labs suddenly needed. Its campus network was built for recruiting, but the same rails can now route PhDs, postdocs, and strong graduates into hourly model training work. That let Handshake add an $80M annualized AI business by August 2025 while the core recruiting business was growing much more slowly.
-
The work changed from cheap clickwork to expert judgment. Earlier RLHF often used large pools of low cost annotators. As labs pushed into math, law, science, and safety evals, they needed people who could actually solve the problem, explain why an answer was wrong, and catch subtle reasoning failures. That favors credentialed networks over generic crowd marketplaces.
-
Handshake already had the trust infrastructure. Universities, career centers, and student profiles gave it a large pre assembled pool of identity checked academic talent. That is different from Mercor’s playbook, which centers on AI screening and matching experts into contracts. Handshake could activate existing supply faster because the relationship with the worker often already existed.
-
The economics also explain why this is attractive even if margins are lower than software recruiting. Handshake’s core job board business carries roughly 80% gross margins, while expert labeling looks more like a marketplace where 60 to 70% of spend passes through to contractors. Even so, it turns a stalled recruiting network into a new revenue stream without rebuilding the supply side from scratch.
The next step is turning this from opportunistic labor supply into infrastructure for human evaluation. If Handshake keeps layering in better quality control, workflow tooling, and deeper profiling of expertise, it can evolve from a campus recruiting network with an AI side business into a durable platform for expert judgment across hiring, research, and model evaluation.