AtlasData unveils privacy-first embeddings service after $40M raise

Tech · 4 min read

AtlasData’s embeddings API emphasizes privacy controls: customers can opt for client-side encryption where embeddings are never visible to AtlasData, or use short-lived embeddings stored only for request duration. The company positioned this as a response to regulatory scrutiny on training data and model leakage.

The product supports hybrid deployment models — a hosted API, private cloud, or on-premises runtime — and integrates with common vector stores and retrieval pipelines. Performance claims include 100ms median query times for 10M vectors on their hosted tier.

The $40 million round, led by Evergreen Capital, will be used to expand compliance certifications, add domain-specific prebuilt embeddings for healthcare and finance, and grow their enterprise sales team. AtlasData also teased an upcoming multimodal embeddings layer for documents and images.