companies

Vectara Inc.

RAG-as-a-service API handling retrieval, reranking, and grounded generation so engineering teams can ship factual AI agents without building a vector pipeline.

Vectara was founded in 2020 by former Cloudera CTO Amr Awadallah along with Amin Ahmad and Tallat Shafaat to package the full retrieval-augmented-generation pipeline, including chunking, embedding, indexing, reranking, and answer generation, behind a single API rather than making teams stitch together a vector database, an embedding model, and an LLM themselves. Its "corpus" abstraction handles multilingual documents natively, and its neural reranker plus Mockingbird model, tuned specifically for RAG-style answer generation, are pitched as reducing hallucination rates compared with calling a general-purpose LLM directly on retrieved chunks. The company is arguably best known outside its paying customers for HHEM, its open-source hallucination detection model and public leaderboard, which has racked up millions of downloads on Hugging Face and become a common reference point whenever people debate which LLMs fabricate facts most often. Vectara raised a Series A in 2024 to expand its grounded-agent pitch toward autonomous, policy-enforced enterprise agents built on the same retrieval core.

Founded

2020

Headquarters

Palo Alto, United States

Sector

infrastructure