Datumo

Selectstar.ai, a Seoul-based company founded in 2019 by Howook Shin and Seyeob Kim, operates a platform for AI training data solutions. Seyeob Kim, who serves as CEO, graduated from KAIST with a degree in electrical and electronic engineering. The company has secured significant funding, including a Series A round of $66.7 million in August 2022, bringing its total funding to $67 million from investors like Kakao Ventures and Company K Partners.

The core of Selectstar.ai's business is to address the challenge of acquiring high-quality data for AI development. It provides a comprehensive suite of services for collecting, labeling, and validating various data types, including images, text, and videos. The business operates on a B2B model, catering to AI companies and research labs that require large volumes of accurate training data. This is facilitated through its crowdsourcing platform, "Cashmission," which engages a large pool of workers to perform data collection and annotation tasks. This platform employs a three-step quality control process involving qualification tests, performance monitoring, and peer-review validation to ensure data accuracy.

The company's platform offers a range of features such as object and line segmentation, classification, transcription, and comparison tools for data labeling. A key technological advantage is its semi-automatic annotation technology, where deep learning models perform initial annotations that are then refined by human workers, reducing cost and time. Furthermore, Selectstar.ai has developed "Datumo Eval," a solution for verifying the reliability of large language models (LLMs) and has been involved in operating AI red team challenges to assess AI model stability. The company also offers an open dataset project, providing free access to datasets built in collaboration with AI companies to foster industry development.

Keywords: AI training data, data labeling, crowdsourcing, data annotation, machine learning data, computer vision data, natural language processing data, dataset collection, data validation, AI data platform, Seyeob Kim, Howook Shin, Kakao Ventures, Cashmission, Datumo Eval, data quality control, AI model evaluation, LLM reliability, object segmentation, text transcription