S
Overview
San Francisco Data is a marketplace for AI production traces, collecting interaction data from AI agents deployed in real-world use. It processes and structures these machine-generated traces and sells them to AI labs to support training, especially reinforcement learning environments that learn from model failures. The company differentiates itself by focusing on real-world interaction data and failure points rather than synthetic labels, drawing on its founders’ experience in scaling data-supply businesses. Its goal is to become the primary supplier of real-world AI interaction data for training advanced AI models.
Work here?Claim your company
Simplify's take
What believers are saying
- Kenan Saleh said revenue jumped from zero to $60M in three months.
- SF Data says it serves 20-plus frontier labs and uses live production failures.
- October 2026 traction makes SF Data a default counterparty for enterprise data monetization.
What critics are saying
- Production traces contain sensitive customer data, creating privacy and contract breach exposure.
- A handful of frontier labs can squeeze prices or replace SF Data with in-house pipelines.
- If access to agent logs dries up, the marketplace loses its core inventory.
What makes San Francisco Data unique
- SF Data packages production traces from real AI work into RL environments.
- Founders Brandon Guo and Faiz Siddiqui built Mercor-style data commercialization experience.
- October 2026 essay and website define a clear Glencore-for-traces positioning.
Help us improve and share your feedback! Did you find this helpful?