
SelectStar Takes Datumo AI Evaluation Platform to Japan
The AMW Read
SelectStar's Tokyo push is an incremental expansion by a data-and-evaluation infrastructure provider into a market where enterprise AI evaluation adoption remains uneven.
SelectStar Takes Datumo AI Evaluation Platform to Japan
SelectStar is bringing its Datumo AI evaluation platform to Tokyo as it targets Japanese enterprises deploying generative AI. The company, founded by six KAIST graduates in 2018, began with AI training-data operations and has expanded into reliability evaluation. Datumo helps customers define service-specific evaluation criteria, then uses LLM-as-a-judge workflows to assess outputs. SelectStar cites deployments with Korean financial institutions, public agencies, LG Electronics, and the Bank of Korea.
The expansion reflects a growing need for evaluation infrastructure as enterprise AI shifts from pilots to customer-facing and regulated workflows. A banking assistant, for example, needs more than fluent answers: it must avoid recommending a competitor's product or departing from approved financial guidance. SelectStar's approach makes the evaluation rubric a consulting and governance exercise before automating recurring tests after model or data changes. That matters in Japan, where prospective customers range from companies already operating AI services to organizations still defining the services they intend to evaluate.
For builders, the practical implication is that evaluation cannot be treated as a one-time launch gate. Teams should establish task-specific criteria, test against them after every model, prompt, retrieval, or data change, and preserve a clear review process for high-impact failures. For investors, the near-term proof point is whether Tokyo pilots convert into reference customers and repeatable deployments through both direct enterprise sales and systems-integration partners. Per the AI Market Watch index, SelectStar has raised about $31M across roughly 5,000 tracked companies, coverage rather than a census.