Vectara

Vectara is a trusted generative AI platform that enables organizations to rapidly build conversational AI applications and semantic search capabilities using retrieval augmented generation (RAG). The company's core offering is RAG-as-a-Service, a serverless platform that combines advanced neural retrieval and large language models to deliver accurate and contextual information to users through natural language interactions.

Vectara's RAG-as-a-Service platform addresses critical challenges in enterprise adoption of generative AI, such as reducing hallucinations, providing explainability and provenance, enforcing access control, enabling real-time updatability of knowledge, and mitigating intellectual property and bias concerns associated with large language models. The platform uses a hybrid search approach, combining keyword search with semantic retrieval and reranking to deliver highly relevant results.

In July 2024, introduced its new Mockingbird LLM designed specifically for RAG applications, engineered to reduce hallucinations and improve structured output. Further, in June 2024, Vectara announced the release of its Multilingual Reranker v1, a state-of-the-art reranker model that supports over 100 languages, enabling impressive zero-shot performance on unseen data and domains across multilingual and cross-lingual settings. This model significantly enhances retrieval performance, outperforming industry leaders and open-source models in benchmarks while offering blazing-fast inference at minimal cost.

Vectara's platform offers various features, including indexing APIs for data ingestion, metadata search filtering, summarizer models, factual consistency scoring for detecting hallucinations, and tools for building UI chat widget frameworks. The company prioritizes user privacy, ensuring that customer data is never used for training models and providing full control over answer history recording.

Key customers and partnerships

In April 2024, Vectara partnered with Datavolo to streamline data gathering and ingestion processes, enabling customers to rapidly develop data pipelines for AI applications using Vectara's platform.