Datumo, a South Korea-based AI startup, has raised $15.5 million to grow its no-code model evaluation tools and expand globally.
Quick Summary – TLDR:
- Datumo raised $15.5 million, led by Salesforce Ventures and others, bringing its total funding to $28 million.
- The company builds AI evaluation tools for safety, bias, and compliance using no-code platforms.
- Its Datumo Eval product automates test data generation and checks for unsafe or biased outputs without coding.
- The funding will fuel R&D and global expansion into Japan, the U.S., and beyond.
What Happened?
South Korean startup Datumo has closed a $15.5 million funding round backed by major investors including Salesforce Ventures, KB Investment, ACVC Partners, and SBI Investment. The company, originally launched as a data labeling service, is now focused on providing no-code AI evaluation tools for enterprises. The raise pushes Datumo’s total funding to $28 million, signaling strong investor confidence as the AI safety race heats up.
From Labeling to Evaluation: Datumo’s Pivot
Founded in 2018 by David Kim and five fellow alumni from the Korea Advanced Institute of Science and Technology, Datumo (formerly SelectStar) started by offering crowdsourced data labeling through a reward-based app. That simple idea turned into a profitable business, generating over $1 million in revenue in its first year, with early support from alumni-led businesses.
However, Datumo soon realized its enterprise customers needed more than labeled data. As clients like Samsung, LG, Hyundai, Naver, and SK Telecom began asking for help scoring and evaluating AI outputs, the company pivoted.
“That’s when we realized: We were already doing AI model evaluation – without even knowing it,” said co-founder Michael Hwang.
AI Safety is Now the Core Mission
Datumo now focuses on building trust, safety, and compliance tools for generative AI. Its flagship product, Datumo Eval, offers a full-stack no-code evaluation platform that automatically generates test datasets and evaluates AI model responses for bias, safety, and correctness. The platform is aimed at non-technical teams, such as those in policy, compliance, or trust and safety, removing the need for manual scripting or deep AI expertise.
The company’s licensed datasets are another standout feature. Unlike some competitors, Datumo uses data sourced from published books, which offer structured human reasoning but require significant cleaning, a challenge the team has embraced.
Industry Context and Strategic Growth
Datumo’s rise comes at a time when AI safety and explainability are top concerns in the enterprise world. A recent McKinsey report highlighted that while 40% of businesses see explainability as a major AI risk, only 17% are actively addressing it. That gap creates a big market opportunity.
The broader AI data and evaluation space has become a hotbed of activity. Meta’s $14.3 billion investment in Scale AI and OpenAI’s decision to stop using Scale’s services have only intensified competition. Datumo sees itself as a blend of Scale AI in dataset provisioning and Galileo or Arize AI in model evaluation, but with a unique no-code, full-stack approach.
The startup also launched Korea’s first benchmark dataset for AI trust and safety, further cementing its leadership in the region.
Global Ambitions with Fresh Funding
Datumo plans to use the new funding to accelerate R&D and develop more automated AI evaluation tools. The startup already employs 150 people in Seoul and recently established a Silicon Valley office in March 2025. Expansion into Japan and the United States is a top priority.
The journey to funding also had a human touch. After hosting Andrew Ng, founder of DeepLearning.AI, for a fireside chat in Korea, CEO David Kim shared the event on LinkedIn. That post sparked interest from Salesforce Ventures, kicking off an eight-month investment conversation.
SQ Magazine Takeaway
I love how Datumo is solving a real pain point in AI right now: How do you trust what the model is doing when you don’t even understand it? What makes their approach stand out is that they are not just building for developers. They are making tools for compliance teams, policy folks, and anyone else in the loop who doesn’t write code but still needs to verify AI behavior. That kind of accessibility is rare in AI tooling. With big-name clients already onboard and a fresh funding boost, Datumo could be one of South Korea’s biggest AI exports.
