The company has raised a Series A led by a regional fund, with two strategic investors participating.
Proceeds go to data collection across dialects that are poorly represented in existing corpora.
The company says it will publish evaluation results rather than headline accuracy claims.