Datacurve

Datacurve

Datacurve is a training-data company that supplies high-quality coding and engineering data to frontier AI model labs. Datacurve sources this data through Shipd, a bounty platform where experienced software engineers earn payouts for completing algorithm, testing and UI challenges, an approach the founders run more like a consumer product than a traditional labeling operation. Founded in 2024 by Serena Ge and Charley Lee, both from the University of Waterloo computer science program and the Y Combinator W24 batch, Datacurve targets the hard, real-world coding tasks that models still struggle with. It also publishes DeepSWE, a long-horizon coding benchmark, to evaluate agentic coding performance.

Production credibility: Datacurve was founded in 2024 by Serena Ge (CEO) and Charley Lee, who met through the University of Waterloo computer science program; Ge previously worked as a machine learning engineer at Cohere, and the pair went through Y Combinator's W24 batch. The company has raised approximately $17.7 million in total: a $2.7 million seed round that included former Coinbase CTO Balaji Srinivasan, followed by a $15 million Series A in October 2025 led by Mark Goldberg at Chemistry, with participation from employees at DeepMind, Vercel, Anthropic and OpenAI. Its Shipd bounty platform has more than 1,400 participating engineers and has paid out over $1 million in bounties for tasks spanning algorithms, testing and UI/UX. Datacurve produces coding challenges, debugging tasks and private-repository benchmarks for foundation model labs, and publishes DeepSWE, a long-horizon coding benchmark. It positions itself as a direct competitor to Scale AI, Surge and Mercor.

Key Features

• Frontier coding and engineering datasets for training and evaluating LLMs
• Shipd: a bounty platform where expert engineers earn payouts for solving coding challenges
• Over 1,400 participating engineers, with $1M+ paid out in bounties
• Debugging tasks and private-repository benchmarks beyond simple labeled data
• DeepSWE: a published long-horizon agentic coding benchmark
• Consumer-product contributor experience rather than a low-cost labeling pipeline
• Coverage of algorithms, testing and UI/UX engineering tasks
• Plans to expand from code into domains such as finance and healthcare

Ideal Use Case

Foundation model labs use Datacurve to buy hard, real-world coding and debugging data, sourced from vetted engineers via Shipd, to improve a model's reasoning and software-engineering ability and to evaluate it on long-horizon coding benchmarks.

How Datacurve differentiates

Against Scale AI, Surge and Mercor, Datacurve narrows its focus to high-difficulty coding and engineering data rather than broad, general-purpose labeling. Where Scale AI operates a very large labeling workforce across many data types, Datacurve runs Shipd as a bounty marketplace that rewards skilled engineers for solving genuine challenges, which is aimed at depth over headcount. Compared with Surge's expert-data positioning, Datacurve leans hard into software engineering specifically and ships its own DeepSWE benchmark. The trade-off is breadth: as an earlier-stage, code-focused company it covers fewer domains than Scale today, though the founders plan to expand into areas like finance and healthcare over time.

FAQ

Q: Who founded Datacurve? A: Datacurve was founded in 2024 by Serena Ge (CEO) and Charley Lee, both connected to the University of Waterloo computer science program. Ge previously worked as an ML engineer at Cohere, and the company went through Y Combinator's W24 batch.

Q: How much funding has Datacurve raised? A: Datacurve has raised about $17.7 million total: a $2.7 million seed round (which included former Coinbase CTO Balaji Srinivasan) and a $15 million Series A in October 2025 led by Mark Goldberg at Chemistry, with participation from employees at DeepMind, Vercel, Anthropic and OpenAI.

Q: What is Shipd? A: Shipd is Datacurve's bounty platform where more than 1,400 software engineers earn payouts for completing coding challenges in areas like algorithms, testing and UI/UX. It has paid out over $1 million in bounties and feeds the high-quality data Datacurve sells to AI labs.

Q: Datacurve vs Scale AI: how do they differ? A: Scale AI is a large, broad data-labeling company spanning many data types. Datacurve focuses specifically on high-difficulty coding and engineering data, sourced from vetted engineers through its Shipd bounty marketplace, and ships its own DeepSWE coding benchmark.

Q: What does Datacurve produce for AI labs? A: Datacurve produces frontier coding datasets, debugging tasks and private-repository benchmarks to help foundation models improve reasoning and software-engineering ability, plus DeepSWE, a long-horizon agentic coding benchmark for evaluation.

tl;dr

Datacurve supplies high-quality coding and engineering training data to frontier AI labs via Shipd, a bounty platform where expert engineers earn payouts for solving challenges. Founded in 2024 by Serena Ge and Charley Lee (YC W24), it has raised about $17.7M and competes with Scale AI and Surge.

Related

Looking for more options? Browse the AI/ML Models directory or read our best AI models listicle. Datacurve is also tracked on Crunchbase.