Proximal Reaches $200M Revenue Using AI-Generated Coding Tasks to Train Models

A San Francisco startup founded just a year ago claims to have reached $200 million in annualized revenue by using AI models to generate synthetic coding tasks—a approach that founders say keeps data production costs substantially below traditional human-labeling methods. Proximal, a 40-person team led by CEO Calvin Chen and co-founder Justus Mattern (both formerly at compute startup Prime Intellect), announced its emergence from stealth alongside a $15 million seed funding round led by General Catalyst, valuing the company at $300 million. The startup supplies high-quality training data specifically for software engineering, and according to Chen's announcement, has achieved profitability.