Datumo raised $15.5M to tackle AI safety with testing and evaluation tools designed for non-tech users. The Seoul startup started as a data-labeling app where anyone could earn money by labeling data. Now it helps companies monitor and improve AI models with no-code tools.
Big Korean firms like Samsung, LG, Hyundai, and SK Telecom rely on Datumo. The company hit $6M revenue in 2024 and counts over 300 clients.
The push came when customers wanted more than labeling—they asked Datumo to score AI outputs and compare results. This revealed Datumo was already doing AI model evaluation without realizing it. CEO David Kim said: 
“We started in data annotation, then expanded into pretraining datasets and evaluation as the LLM ecosystem matured.”
Datumo’s platform, Datumo Eval, automatically generates test data and finds unsafe, biased, or wrong AI responses without manual scripting. This sets it apart from peers like Scale AI or Arize AI. It also uses licensed datasets crawled from books, bringing structured human reasoning hard to get elsewhere.
The company’s latest $15.5M raise includes Salesforce Ventures, KB Investment, ACVC Partners, and SBI Investment, pushing total funding near $28M. Kim credits a LinkedIn share of a fireside chat with Andrew Ng for attracting investors.
“Following several meetings and Zoom calls, the investors extended a soft commitment. The entire funding process took about eight months,” said co-founder Michael Hwang.
Datumo plans to use the cash for R&D on automated evaluation tools and to expand globally in South Korea, Japan, and the US. The startup opened a Silicon Valley office in March and employs 150 people in Seoul.
The raise comes on the heels of Meta’s $14.3B investment in Scale AI, highlighting the war for trustworthy AI training data. However, Meta recently stopped using Scale AI’s services, signaling a shifting landscape in AI data sourcing.
TechCrunch event
San Francisco | October 27-29, 2025
Image Credits: Datumo
