Handshake AI's Early Metrics Focus

Handshake's AI Data Labeling Business: From Zero to $100M+ in Under a Year

Handshake transformed its decade-old college recruiting platform into one of the fastest-growing AI businesses by leveraging its network of 18 million students and experts to provide high-quality training data to frontier AI labs.

The Strategic Opportunity in AI Data Labeling

AI model training has shifted from pre-training (ingesting all available internet data) to post-training (improving specific capabilities with expert data)
The market evolved from needing generalist data labelers to requiring domain experts:
- "The models have gotten so good that the generalists are no longer needed. What they really need is experts across every area that the models are focused on."
- Labs are targeting economically valuable capability areas: advanced STEM, science, math, law, medicine, finance
Handshake's unique advantage: "The only moat in human data is access to an audience"
- 18 million professionals including 500,000 PhDs and 3 million master's students
- Established trust relationship with universities and students
- Zero customer acquisition cost compared to competitors spending "tens of millions on performance advertising"

Types of AI Training Data Being Created

Breaking models in advanced domains and providing correct answers
- "In order to break a model...if you're a PhD in physics, you can go in multiple subdomains of physics and prove where the model's actually breaking"
Step-by-step reasoning processes for complex problems
- "They're really focused on the steps to get there...like 10 steps in a math problem, steps 6-10 are wrong"
Trajectories: recording complete problem-solving processes
- "A trajectory is basically the entire environment that is collecting what you're doing - your screen, your mouse"
Rubrics for non-verifiable domains
- "Models can sit in the middle as a judge and actually understand what is a good educational design or a good MRI result"

Building a New Business Inside an Established Company

Complete separation of teams and responsibilities
- "Separate engineering team, separate design team, separate accounts and operations team, separate finance team"
- "People only had one job and one job only - making Handshake AI successful"
- Separate all-hands meetings, onboarding, and recruiting
Founder-led execution with direct involvement
- "I focused 80+ percent of my time and attention on just this"
- "I was pretty hands-on...everyone reported directly to me"
- Hired team members with entrepreneurial experience comfortable with ambiguity
Different operating cadence and culture
- "We're a lot more metrics oriented...way more focused on operating with data and metrics and rigor from an early stage"
- Created urgency: "Leave nothing to chance...how do you make sure three months from now, six months from now, you have no regrets?"
- Set different expectations: "This is a 24/7 job...this is an early-stage company"
- "A huge celebratory culture...calling out the people that are putting up points and creating a really fun environment around impact"

Execution Principles for Rapid Growth

Focus on quality first before scaling
- "Make sure that we could deliver high quality data to one customer before we expand to anyone else"
- Built internal post-training team to verify data quality
Create an expert-first experience
- "PhD students expect to be treated different than lower-cost international labor"
- Built community, cohort-based training, and instructional design
Optimize for retention and lifetime value
- "LTV is calculated pretty simply in this business...based on the retention of a person and how many projects they can participate in"
- Treating experts well leads to higher retention rates and project participation
Maintain urgency and momentum
- "There will never be a time like this...where there's unlimited demand"
- "Get on a plane to go talk to a customer, make the late night push, check the data six times over again"