From Labeling Side Hustle to AI Evaluation Powerhouse
Datumo, founded in 2018 by CEO David Kim and five fellow alumni from KAIST, began its journey as a data labeling company under the name SelectStar. Frustrated by the inefficiencies in traditional labeling processes, Kim introduced a reward-based mobile app that allowed users to label data in their spare time.
“We started in data annotation, then expanded into pretraining datasets and evaluation as the LLM ecosystem matured,”
— David Kim, CEO of Datumo
Even before their platform was fully operational, the team had secured tens of thousands of dollars in pre-contract sales — largely from fellow KAIST alumni startups.
Datumo’s client list quickly grew to include South Korean giants like:
- Samsung and Samsung SDS
- LG Electronics and LG CNS
- Hyundai
- Naver
- SK Telecom
By 2024, the company had over 300 clients and generated $6 million in revenue.
The Pivotal Shift: Listening to the Market
The company’s evolution came naturally when clients began asking for more than just labeled data. They needed AI model evaluation — comparing outputs, testing for safety and fairness, and more. Co-founder Michael Hwang recalls:
“That’s when we realized: We were already doing AI model evaluation — without even knowing it.”
— Michael Hwang, Co-founder of Datumo
This realization pushed Datumo to release South Korea’s first benchmark dataset dedicated to AI trust and safety, firmly establishing their presence in the growing world of LLM evaluation tools.
Why Datumo Stands Out in a Crowded Market
As AI faces increasing scrutiny, companies are racing to ensure safe and explainable models. Datumo is positioning itself uniquely:
- Their flagship product Datumo Eval is a no-code evaluation platform.
- Designed for non-technical users such as policy, compliance, and trust & safety teams.
- Automatically generates test data and evaluations to detect bias, unsafe or incorrect outputs (like AI hallucinations) — no manual scripting needed.
- Offers automated red teaming, dashboard insights, and licensed datasets drawn from published books — a rich, structured, but notoriously difficult data source.
“Our datasets go beyond the typical. They contain deep human reasoning — but they’re hard to clean, and that’s our edge.”
— David Kim, CEO of Datumo
Timing the Market: Riding the Scale AI Shake-Up
Datumo’s rise coincides with major disruption in the space. In June 2025, Meta made headlines with a $14.3 billion investment in Scale AI, acquiring a 49% stake. The move triggered skepticism and fallout:
- OpenAI, Google, and Microsoft reportedly cut ties with Scale over neutrality concerns.
- Scale AI laid off 200 employees (14% of its workforce).
- Analysts questioned if Meta overpaid amid performance issues.
This turbulence cracked open a market opportunity — one Datumo seems ready to seize.
According to a McKinsey report:
- 40% of organizations see AI explainability as a major risk.
- Only 17% are actively addressing it — leaving a huge gap.
Fuel for Growth: Backed by Salesforce and More
To capitalize on this moment, Datumo has raised $15.5 million, bringing its total funding to $28 million. The round was led by Salesforce Ventures, alongside KB Investment, ACVC Partners, and SBI Investment.
Interestingly, the relationship with Salesforce Ventures began after CEO Kim posted about a fireside chat with Andrew Ng on LinkedIn — catching the firm’s eye.
“It took about eight months and multiple Zoom calls, but they were in,”
— Michael Hwang, Co-founder
Looking Ahead: R&D and Global Expansion
Datumo plans to use the fresh capital to:
- Accelerate R&D into automated AI evaluation tools.
- Strengthen presence in Japan, South Korea, and the U.S..
- Scale operations from its Silicon Valley office established in March 2025.
- Continue hiring and advancing enterprise-ready AI safety tech.
The global AI platform market is projected to hit $366.93 billion by 2031, growing at a 31.41% CAGR — and Datumo is poised to ride that wave.
