Most organizations claim that they are not fully prepared to make use of generative artificial intelligence in a secure and responsible manner, According to the last McKinsey report. One of the problems is an explanation – understanding how and why AI makes some decisions. While 40% of respondents consider it a significant risk, only 17% actively deal with it, in keeping with the report.
Based on Seoul Data It began as a company labeling AI and now desires to help corporations build safer artificial intelligence with tools and data that enable testing, monitoring and improving their models – without the requirement of technical knowledge. On Monday, the startup collected $ 15.5 million, which is able to bring her sum to around USD 28 million, from investors, including Salesforce Ventures, KB Investment, ACVC Partners and SBI Investment.
David Kim, the general director of Dataumo and former AI researcher from Korea, defense development agency, was frustrated with a time -consuming character of knowledge labeling, so he got here up with a recent idea: an application based on a prize, which allows everyone to mark data in free time and earn money. Startup confirmed the idea at the startup competition in Kaist (Korea Advanced Institute of Science and Technology). Who was the co -founder of Datumo, previously often known as Selectstar, along with five Kaist graduates in 2018.
Even before the application was fully built, Dataumo secured tens of 1000’s of dollars of sales against the contract during the phase of discovering customers, mainly from corporations and startups managed by Kaist Alumni.
In the first 12 months, the startup exceeded $ 1 million in revenues and secured several key contracts. Today, the startup counts the major Korean corporations akin to Samsung, Samsung SDS, LG Electronics, LG CNS, Hyundai, Naver and Seul Giant SK among its clients. However, a few years ago, customers began to ask the company to go beyond easy data labeling. The 7-year startup currently has over 300 customers in South Korea and generated about $ 6 million revenues in 2024.
“They wanted us to get our outputs of the AI model or compare them with other outputs,” said Techcrunch Michael Hwang, co-founder of Datumo. “Then we realized: we have already carried out the AI model – without even knowing it.” Hwang added that Datumo doubled in this area and released the first in Korea in Korea Benchmark DataSet, which focused on AI trust and safety.
“We started with data annotation, and then we extended to pretring data sets and an evaluation as the LLM ecosystem matured,” said Kim Techcrunch.
TechCrunch event
San Francisco
|.
October 27-29 2025
Recent meta investments in the amount of $ 14.3 billion at the AI scale of the company dealing in data marking emphasizes the importance of this market. Shortly after this offer, the manufacturer of the AI and Meta competitor OPENAI stopped using the AI Scale services. The meta offer also signals that the competition for training and training is intensifying.
Datumo shares some similarities with corporations akin to AI Scale AI in the scope of sharing data sets, in addition to with Galileo and Aize AI in AI assessments and monitoring. However, it stands out through licensed data sets, in particular data crawled from published books, which in keeping with the company offers a wealthy human reasoning, but is extremely difficult to scrub, in keeping with Kim.
Unlike its peers, Datumo also offers a full rating platform called Data evaluationwhich routinely generates test data and assessments to ascertain dangerous, biased or incorrect answers without the need for manual script, she added Kim. The Signature product is a tool for evaluation without a code designed for non-programmers, akin to those in terms of principles, trust and security and compatibility teams.
When asked about attracting investors, akin to Salesforce Ventures, who explained that the startup had previously hosted Fireside chat with Andrew Ng, the founding father of Deeplearning.Ai, at a party in South Korea. After the event, Kim shared a session at LinkedIn, which drew the attention of Salesforce Ventures. After several meetings and mixtures, investors expanded their soft commitment. Hwang said that the entire financing process lasted about eight months.
New funds shall be used to speed up research and development activities, especially in the development of automated assessment tools for AI Enterprise, and for scaling global operations on the market in South Korea, Japan and the United States, which has 150 employees in Seoul, also established presence in the Silicon valley.