Seoul-Based Datumo raises $15.5M to take on Scale AI in LLM Evaluation and Safety

Datumo, a Seoul-based AI model evaluation and data labeling company, has raised $15.5 million in funding from Salesforce Ventures, KB Investment, ACVC Partners, SBI Investment, and others. The round brings its total funding to about $28 million.

Founded in 2018 by David Kim (ex–AI researcher at Korea’s Agency for Defense Development) and five KAIST alumni, Datumo started as SelectStar, a reward-based crowdsourcing app for data labeling. It has since evolved into a full-stack AI evaluation platform used by Samsung, LG Electronics, Hyundai, Naver, and SK Telecom, serving 300+ clients and generating $6M in 2024 revenue.

From labeling to trust and safety

The pivot came when enterprise clients began asking Datumo to score AI model outputs and benchmark performance. Datumo launched Korea’s first AI trust and safety benchmark dataset and now offers:

• Licensed datasets from published books, offering rich structured reasoning data

• Datumo Eval, a no-code evaluation tool for non-developers in policy, compliance, and trust & safety teams

• Automatic generation of test data to detect unsafe, biased, or incorrect LLM responses

Market positioning

Datumo operates at the intersection of Scale AI (pretraining datasets) and Galileo/Arize AI (evaluation and monitoring), but differentiates with proprietary licensed datasets and its non-developer-friendly evaluation tools.

The company’s funding comes amid intensifying competition for AI training and evaluation data, highlighted by Meta’s $14.3B investment in Scale AI and OpenAI’s shift away from using Scale’s services.

Growth and expansion

The new capital will fund:

• R&D on automated AI evaluation tools for enterprises
• Expansion into Japan and the United States
• Scaling go-to-market efforts beyond South Korea

Datumo currently employs 150 people and established a Silicon Valley presence in March 2025.

“We started in data annotation, then expanded into pretraining datasets and evaluation as the LLM ecosystem matured,” said Kim. “Our mission is to help enterprises deploy safer, more reliable AI without needing deep technical expertise.”