The latest McKinsey report shows that most companies are still not prepared to use generative AI safely and responsibly. One major concern is explainability - understanding how and why AI makes specific decisions. While 40% of respondents consider this a major risk, only 17% of companies are actively addressing it.
Based in Seoul, Datumo started as an AI data annotation company and is now focused on providing tools and data for testing, monitoring, and improving AI models, making them more secure, without requiring technical expertise. This Monday, the company announced a $15.5 million funding round, bringing its total funding to about $28 million. Investors include Salesforce Ventures, KB Investment, ACVC Partners, and SBI Investment, among others.
David Kim, CEO of Datumo, previously worked as an AI researcher at the Korean Agency for Defense Development. Frustrated by the tedious and time-consuming data annotation process, he came up with a new idea: a reward-based application that allows anyone to annotate data during their free time and earn money. This idea was validated in a startup competition at KAIST. In 2018, David Kim co-founded the company with five KAIST alumni, initially named SelectStar.
Before the application was fully developed, Datumo secured several pre-orders worth tens of thousands of dollars during the customer discovery phase of the competition, with clients mostly being companies and startups founded by KAIST alumni. The company achieved over one million dollars in revenue in its first year and signed multiple significant contracts. Today, Datumo's clients include large Korean enterprises such as Samsung, Samsung SDS, LG Electronics, LG CNS, Hyundai, Naver, and SK Telecom, a major telecommunications company in Seoul. As of 2024, the company has over 300 clients and annual revenue of approximately six million dollars.
A few years ago, clients began asking Datumo not only to provide data annotation but also to score AI model outputs or compare them with other outputs. Michael Hwang, a co-founder, said this made them realize they were already doing AI model evaluation. Subsequently, the company increased its investment and launched South Korea's first benchmark dataset focused on AI trust and security.
David Kim stated that Datumo started with data annotation, expanded its business to pre-training datasets and model evaluation as the large language model ecosystem developed, and gradually positioned itself as a provider of AI safety and trust solutions.