Keep pulling the thread on Garrett Lord.
Handshake launched a new AI data labeling business in January that reached $50 million in Annual Recurring Revenue (ARR) within four months.
Handshake's new AI data business is projected to surpass $100 million in Annual Recurring Revenue (ARR) within its first 12 months of operation.
The majority of performance improvements in frontier AI models are now derived from post-training processes rather than pre-training.
The AI data labeling market has shifted from needing generalist, low-cost labor to requiring subject matter experts because models have become too advanced for generalists to improve.
Garrett Lord asserts that the only sustainable moat in the human-generated AI data market is access to a large, engaged audience.
Handshake's AI data business is working with seven of the frontier AI labs.
Handshake's core job marketplace business generates approximately $200 million in annual revenue.
There is currently unlimited demand for high-quality, high-volume AI training data, allowing providers to sell nearly everything they can produce.
Handshake's platform is used by every Fortune 500 company, over 1,500 colleges, over 20 million students and alumni, and over 1 million companies for hiring graduates.
Gains from pre-training AI models began to asymptote approximately 18 to 24 months ago as labs had largely ingested all available knowledge from the internet.
Handshake's network consists of 18 million professionals, including 500,000 PhDs and 3 million master's students.
Handshake hired a former post-training lead from Meta to build its internal research and data quality assessment team.