Seoul-based artificial intelligence evaluation company Datumo has secured $15.5 million in new funding, bringing its total capital raised to approximately $28 million as it seeks to compete with established players like Scale AI in the model assessment market.
The funding round was led by Salesforce Ventures alongside KB Investment, ACVC Partners, and SBI Investment. The investment signals growing investor confidence in AI safety and evaluation tools as enterprises increasingly deploy large language models across their operations.
From Crowdsourcing to Enterprise AI Safety
Founded in 2018 by former defense researcher David Kim and five graduates from Korea Advanced Institute of Science and Technology, Datumo originally operated under the name SelectStar. The company initially focused on reward-based crowdsourcing for data annotation tasks.
The transformation came when enterprise customers began requesting more sophisticated services beyond basic data labeling. “They wanted us to score their AI model outputs or compare them to other outputs” explained co-founder Michael Hwang. “That’s when we realized: We were already doing AI model evaluation — without even knowing it.”
This client demand prompted Datumo to pivot toward comprehensive AI model assessment, eventually launching Korea’s first benchmark dataset focused on AI trust and safety protocols.
Building a No-Code Evaluation Platform
Datumo’s current offerings center around three core products. The company provides licensed datasets sourced from published books, which CEO Kim describes as offering rich structured human reasoning data that proves challenging to clean and process.
The startup’s flagship product, Datumo Eval, represents a no-code evaluation platform designed specifically for non-technical teams in policy, compliance, and trust and safety departments. The tool automatically generates test data and assessments to identify unsafe, biased, or incorrect responses from language models without requiring manual scripting.
This approach differentiates Datumo from competitors by making AI evaluation accessible to business users who lack technical programming skills but need to assess model performance and safety.
Serving Korea’s Tech Giants
The company has built a substantial client base exceeding 300 organizations, including major Korean corporations such as Samsung, LG Electronics, Hyundai, Naver, and SK Telecom. Datumo generated approximately $6 million in revenue during 2024, demonstrating significant commercial traction.
The startup’s growth trajectory began early, with the company surpassing $1 million in revenue during its first year of operations. Initial sales came largely through connections from the KAIST alumni network, providing crucial early validation for the business model.
Expanding Beyond Korea
With 150 employees based in Seoul, Datumo established its first international presence in Silicon Valley in March, signaling ambitions for global expansion. The new funding will support research and development efforts focused on automated evaluation tools for enterprise AI applications.
The company plans to scale operations across South Korea, Japan, and the United States, targeting the growing demand for AI safety and evaluation services among enterprise customers.
The funding process itself took approximately eight months to complete, beginning after CEO Kim shared content featuring DeepLearning.AI founder Andrew Ng on LinkedIn, which attracted attention from Salesforce Ventures. “We started in data annotation, then expanded into pre-training datasets and evaluation as the LLM marketplace matured” Kim noted.
As enterprises continue deploying AI models at scale, Datumo’s focus on making evaluation tools accessible to non-technical teams positions the company to capture market share in the expanding AI safety sector.