Datacurve

About Datacurve

Datacurve is a platform that produces high-quality coding data for foundation model companies and generative AI dev-tool startups. Founded in 2024 by Serena Ge and Charley Lee, the company emerged from the Y Combinator Winter 2024 batch to address the bottleneck of acquiring high-quality training data for AI models. Both founders are University of Waterloo computer science dropouts. Ge, the CEO, previously interned at Cohere, working on LLM reasoning capabilities, while Lee, the CTO, interned at Google. Their shared background in AI research groups and building previous projects together led to the creation of Datacurve.

The company's business model centers on selling curated, expert-quality code datasets to its clients. Datacurve operates a gamified, bounty-based platform called Shipd that attracts and incentivizes highly skilled software engineers to solve complex coding problems, debug tasks, and perform code reviews. This approach allows the firm to generate proprietary, research-grade datasets that are not easily obtained through scraping or synthetic generation. Contributors, including engineers from top tech firms like DeepMind and OpenAI, are paid for their outputs, with the company having distributed over $1 million in bounties. The primary clients are foundation model labs that need expert data to improve general model capabilities like code completion and debugging, and AI developer tool startups requiring custom data for specific tasks such as UI generation or framework-specific code optimization.

By October 2025, Datacurve had raised approximately $17.7 million in total funding, including a $15 million Series A round led by Chemistry. This followed a $2.7 million seed round. The San Francisco-based company aims to solve the critical need for nuanced, post-training data required for advanced AI models to improve their reasoning and real-world performance. Initially focused on coding data, Datacurve plans to expand into other sectors like finance and healthcare.