Many organizations admit they lack readiness to utilize generative AI in a safe and responsible manner, as per a recent McKinsey report. A key concern is explainability – understanding the rationale behind AI decisions. While 40% of respondents see it as a significant risk, only 17% are actively working on it, according to the report.
Seoul-headquartered Datumo, initially a provider of AI data labeling services, is now focusing on assisting businesses in constructing safer AI models using tools and data for testing, monitoring, and enhancing their models – all without the need for technical expertise. Recently, the startup secured $15.5 million in funding, bringing its total raised to around $28 million. Notable investors in this round include Salesforce Ventures, KB Investment, ACVC Partners, and SBI Investment.
David Kim, the CEO of Datumo and a former AI researcher at Korea’s Agency for Defence Development, recognized the tedious nature of data labeling and devised a solution: a reward-based app that allows individuals to label data during their free time and earn money. The concept was validated at a startup competition at KAIST (Korea Advanced Institute of Science and Technology). Kim, alongside five KAIST alumni, co-founded Datumo, formerly known as SelectStar, in 2018.
Even before the app’s full development, Datumo secured significant pre-contract sales during the customer discovery phase of the competition, primarily from businesses and startups led by KAIST alumni. In its inaugural year, the startup surpassed $1 million in revenue and secured crucial contracts. Today, Datumo’s clientele includes major Korean corporations like Samsung, LG Electronics, Hyundai, Naver, and SK Telecom. The company, founded seven years ago, now boasts over 300 clients in South Korea and generated approximately $6 million in revenue in 2024.
Michael Hwang, co-founder of Datumo, revealed that clients began requesting more than just data labeling services, prompting the company to delve into AI model evaluation. This led to the release of Korea’s first benchmark dataset focused on AI trust and safety.
Kim mentioned, “We began with data annotation, then expanded into pretraining datasets and evaluation as the LLM ecosystem matured.”
Techcrunch event
San Francisco
|
October 27-29, 2025
Image Credits:Datumo
Meta’s recent investment of $14.3 billion in data-labeling company Scale AI underscores the importance of this sector. Following this deal, AI model maker OpenAI, a Meta competitor, ceased utilizing Scale AI’s services, signaling heightened competition for AI training data.
Datumo, similar to Scale AI in pretraining dataset services and akin to Galileo and Arize AI in AI evaluation and monitoring, stands out through its licensed datasets, particularly data sourced from published books. This dataset, according to CEO Kim, offers rich structured human reasoning but poses challenges in terms of cleanliness.
Datumo differentiates itself by offering a comprehensive evaluation platform called Datumo Eval, which automates test data generation and evaluations to detect unsafe, biased, or erroneous responses without manual scripting. The no-code evaluation tool caters to non-developers, such as those in policy, trust and safety, and compliance roles.
Kim shared the journey of attracting investors like Salesforce Ventures, citing a fireside chat with Andrew Ng, founder of DeepLearning.AI, in South Korea as the starting point. The session, shared on LinkedIn, caught the attention of Salesforce Ventures, leading to fruitful discussions and eventual funding after eight months.
The fresh funding will fuel R&D efforts, especially in creating automated evaluation tools for enterprise AI, and expand global operations across South Korea, Japan, and the U.S. With 150 employees in Seoul, Datumo also established a foothold in Silicon Valley earlier this year.
Do you have feedback for us? Help shape TechCrunch’s coverage and events by sharing your thoughts through this survey for a chance to win a prize!
