Seoul’s Datumo Secures $15.5M from Salesforce to Rival Scale AI
▼ Summary
– Most organizations feel unprepared to use generative AI responsibly, with only 17% addressing the key risk of AI explainability despite 40% recognizing its importance.
– Datumo, a Seoul-based AI startup, raised $15.5 million to expand its tools for safer AI model testing and monitoring, bringing total funding to $28 million.
– Founded in 2018, Datumo began as a data labeling company but evolved into AI model evaluation after client demand, now serving major Korean firms like Samsung and LG.
– Datumo differentiates itself with licensed datasets and a no-code evaluation platform, Datumo Eval, designed for non-technical teams to detect biases or errors in AI outputs.
– The startup plans to use its new funding for R&D and global expansion, targeting South Korea, Japan, and the U.S., after establishing a Silicon Valley presence in March.
Businesses worldwide struggle with implementing generative AI responsibly, with only 17% actively addressing critical concerns like explainability, according to recent industry research. This gap in AI safety and transparency presents a growing market opportunity that Seoul-based startup Datumo aims to address through its innovative platform.
The company recently secured $15.5 million in Series A funding, with participation from prominent investors including Salesforce Ventures, bringing its total funding to $28 million. Founded in 2018 by former defense AI researcher David Kim and five KAIST graduates, Datumo initially focused on crowdsourced data labeling before expanding into comprehensive AI model evaluation tools.
Datumo’s journey began with an unconventional approach – a mobile app that allowed everyday users to label data for compensation. This concept gained immediate traction, generating $1 million in first-year revenue and attracting major Korean corporations like Samsung, LG, and Hyundai as clients. As client needs evolved beyond basic data annotation, the company recognized its natural progression into AI model assessment.
“We discovered we were already performing model evaluations without realizing it,” explained co-founder Michael Hwang. This realization led to the development of Korea’s first benchmark dataset focused on AI trust and safety, positioning Datumo at the forefront of responsible AI development.
The company now serves over 300 clients and generated approximately $6 million in revenue last year. Its technology stack includes proprietary licensed datasets featuring book-derived content that provides uniquely structured human reasoning patterns. The platform’s flagship Datumo Eval tool offers automated testing for bias, safety, and accuracy without requiring coding expertise, making it accessible to compliance and policy teams.
The recent funding round follows growing industry recognition of data quality’s importance, highlighted by Meta’s $14.3 billion investment in Scale AI. While Datumo shares some similarities with competitors in data provisioning and evaluation, it differentiates itself through its specialized datasets and full-stack evaluation capabilities.
Kim attributes the successful fundraising to strategic networking, including a fireside chat with AI pioneer Andrew Ng that caught Salesforce Ventures’ attention. The eight-month funding process will now fuel global expansion, with new offices in Silicon Valley and plans to strengthen presence in Japan and the U.S. market.
With 150 employees currently based in Seoul, Datumo represents South Korea’s growing influence in the global AI ecosystem. The company’s evolution from data labeling to comprehensive model evaluation mirrors the industry’s shifting priorities toward trustworthy, explainable AI systems.
(Source: TechCrunch)
