Handshake

Handshake had been quietly building one of the world's largest expert networks for a decade. With 18 million professionals, including 500,000 PhDs and 3 million master's students, they'd become the de facto job platform for college grads and young professionals. But as Garrett Lord and his team watched the AI landscape shift in late 2023, they noticed something critical: frontier AI labs were reaching out directly to Handshake, trying to recruit their experts for data labeling work. These labs had realized that training models required not generalists doing simple tasks (like drawing bounding boxes), but credentialed experts—PhD physicists, biologists, educators, mathematicians—to break models, create step-by-step reasoning chains, and generate the sophisticated training data that actually moved the needle.

Middleman companies in the data labeling space were struggling. They were spending tens of millions monthly on Google Ads, Instagram, and LinkedIn, deploying 200+ recruiters to acquire individual experts. The user experience was transactional and misaligned with what PhDs expected. Garrett realized Handshake had the unfair advantage everyone was chasing: a trusted audience, zero customer acquisition costs due to university partnerships, and a brand that PhDs already knew and trusted.

Over Christmas and New Year 2023-2024, Garrett started the exploration. He flew around meeting frontier lab leaders, understanding what they actually needed. By January 2024, he assembled a team of people from the human data world and started building. The philosophy was to build an expert-first platform, not another generic crowdsourcing experience. They created community features, built instructional design to teach PhDs how to use their tools, set up training cohorts with peer groups, and established clear swim lanes where they could pre-build high-quality data, validate it with their own post-training team and GPUs, and then sell packaged datasets to multiple labs.

Crucially, they treated experts like experts. A PhD biologist making $150-200/hour creating novel training data had entirely different expectations than someone paid $25/hour as a teaching assistant. Handshake invested in the community, the training, and the experience.

The customers found them. Seven of the world's frontier AI labs—essentially every lab building the best large language models—started using Handshake's platform. There was no cold outreach needed. The demand was unlimited. Garrett recalled: "There will never be a time like this. I've never seen anything like it. I doubt I'll ever feel anything like this in business again, where there's unlimited demand."

The direct relationships with labs, combined with their ability to convert experts from their existing network at exceptional rates, created immediate traction. Handshake's 10 years of building trust with students meant these PhDs and master's students converted at rates far exceeding what competitors achieved through expensive ad campaigns.

The structural advantages were immense. Other players had 200 recruiters, spent millions on ads, and churned users quickly. Handshake had zero CAC—they could reach experts through their existing partnerships with 1,600 universities (92% of top 500 schools). Their LTV was high because retention and repeat engagement were strong; experts trusted the platform, stayed engaged, and took multiple projects. They built their own post-training team to validate data quality, renting GPUs to ensure each unit of data actually improved model performance.

What didn't work elsewhere: treating PhDs like generalists, poor onboarding, transactional payment systems, high churn. Handshake did the opposite.

In 4 months, Handshake's data labeling business hit $50 million in revenue. By month 8, it's on pace to exceed $100 million ARR—essentially matching the revenue of their core business in just a fraction of the time. Garrett expects they'll "blow through" the $100M number. The business has become a second company operating inside the first, requiring different scaling muscles but drawing on the same fundamental asset: the largest expert network in the world.

The larger story resonates beyond just numbers. Young people trained on frontier AI, becoming paid researchers while still in school, learning cutting-edge techniques, and building portfolios that will define their careers. The models improve. The economy evolves. And a platform built for one purpose—connecting students to jobs—became indispensable for the next frontier of AI itself.