DatologyAI

DatologyAI describes itself as a data refinery for AI model builders. It runs on a customer's own data and does not sell data or source tokens; instead it improves the customer's data. Its four-stage pipeline — Clean, Curate, Create, Compose — produces staged training datasets for mid-training on open models or pre-training of foundation models, deployed on the customer's private cloud. The company positions data quality as a compute multiplier and states its mission is to democratize data curation so that every company, not just those with big budgets and specialized teams, can train better models faster and more affordably.

DatologyAI was founded to help AI models select the best data to be trained on, according to its about page. The company states that it democratizes data curation, allowing every company to easily train its own custom model on the right data without needing to invest massive resources. Its platform leverages cutting-edge research to identify redundant, noisy, or otherwise harmful data points and to manage the entire process from data in blob storage to the dataloader used for training code.

Data quality as the biggest lever in AI and a compute multiplier; fully automated, scalable data curation that is easy to implement.

• Data refinery pipeline

  A four-stage pipeline: Clean (remove poorly formatted, empty, short, and eval-contaminated data), Curate (tailor data by quality, task, and taxonomy), Create (generate synthetic data to maximize task-relevance and diversity), and Compose (mix data sources optimally to produce staged training datasets).

1. 1Datology: Train Better Models, Faster and Smallerdatologyai.com
2. 2Our Mission: Democratize AI Data Curation | DatologyAIdatologyai.com