AI training unicorn Snorkel AI just laid off 13% of its entire workforce (www.businessinsider.com)

Snorkel AI, a $1.3B Stanford spinout that helps companies generate and curate training data, cut 31 of its roughly 240 employees (about 13%) as it shifts focus toward a "data-as-a-service" model. The layoffs, confirmed to Business Insider, hit software engineering hardest (13 engineers), while most AI-facing roles were spared: none of the applied AI engineers or research scientists were cut and only 3 of 25 people with "AI" in their titles were let go. Snorkel raised a $100M Series D in May and lists partners like Google and Anthropic, but says it is deprioritizing legacy parts of the business to concentrate resources where it can better serve customers. For the AI/ML community this is a signal about market dynamics in the data-labeling and human-in-the-loop space: startups are consolidating and pivoting to higher-margin, productized data services as demand and unit economics evolve. Preserving core research and applied AI hires suggests Snorkel is prioritizing model/data quality and platform capabilities over broader engineering expansion, but trimming software teams could slow feature development or integrations. The move echoes recent cutbacks at Scale AI and underscores that even well-funded players in the training-data ecosystem must tightly align org structure with profitable workflows and enterprise customers.