DatologyAI
Data refinery that curates training data for AI model builders
About
DatologyAI (Datology) builds a data curation platform that turns an organisation's proprietary, public and licensed data into high-quality training datasets for mid-training open models or pre-training custom foundation models, deployed in the customer's own private cloud. Its pipeline cleans, curates, generates synthetic data and composes multi-stage training mixes, with the goal of training faster, cheaper and smaller models. It was co-founded by CEO Ari Morcos (formerly Meta FAIR and DeepMind), CTO Bogdan Gaza and Matthew Leavitt (formerly head of data research at MosaicML). Published work includes a legal-domain model built with Thomson Reuters, the DataSmith autonomous data-research harness, and the open-source Zephon data loader released in October 2026.
- Website
- datologyai.com
- Added via
- web
- Ownership
- unclaimed
Where to go
Products
- Datology Data Curation Platform — Clean, curate, create and compose training data inside your own environment · Saas
- Zephon — Open-source, elastically deterministic data loader for foundation model training · Open Source