DatologyAI builds automated data curation technology for training deep learning models, founded in 2023 by CEO Ari Morcos alongside Bogdan Gaza and Matthew Leavitt. The team draws from Meta AI, DeepMind, and MosaicML, and has raised $57.6 million across an $11.65 million seed round led by Amplify Partners and a $46 million Series A extension.
Rather than manually filtering training data, DatologyAI's modality-agnostic algorithms identify redundant, noisy, or otherwise unhelpful data points across text, image, and other data types without needing labels, letting AI labs train comparable or better models on a smaller, cleaner dataset. The pitch to model developers is faster training runs and lower compute cost for a given level of model quality, rather than a marginal accuracy bump.
DatologyAI targets AI labs and enterprises training their own foundation or domain-specific models at a scale where data curation meaningfully affects training cost. There is no published self-serve pricing; the company works with customers through direct sales given the technical, project-specific nature of large-scale data curation engagements.
