Enterprise Deep Research: The Next Killer App for Enterprise AI — Ofer Mendelevitch, Vectara
AI Engineer Code 2025 · 5:19
← All organizationswww.vectara.com ↗
Enterprise AI agents and retrieval-augmented generation
Vectara
Vectara provides an API-first platform for enterprises building AI agents and assistants grounded in their own data. Developers can combine retrieval across text, documents, and images with generation, conversational context, and tool use. The platform brings retrieval, orchestration, and observability together, with audit trails, access controls, and factual-consistency checks to help teams build and govern applications without assembling each part themselves.
Founded in 2020 by Amr Awadallah, Amin Ahmad, and Tallat Shafaat, Vectara is now led by co-founder and CEO Tallat Shafaat. Its research contributions include the Hughes Hallucination Evaluation Model (HHEM), released openly in 2023 to assess whether generated text is supported by its input. This measures consistency with supplied evidence, rather than universal truth. An improved version powers Vectara’s Factual Consistency Score; the company reported more than 100,000 downloads of HHEM by April 2024.
Vectara supports SaaS, VPC, on-premises, and air-gapped deployment, giving organizations choices over where their data and agents operate. Enterprise applications extend beyond internal knowledge search into failure analysis, contract lifecycle management, and manufacturing workflows. In 2026, the company named Sandisk and Altera as new customers and described Broadcom’s use in customer experience and silicon engineering.
3 talks
Newest first
open-rag-eval: RAG Evaluation without "golden" answers.
AI Engineer World's Fair 2025 · 5:03
The Hidden Costs of Building Your Own RAG Stack — Ofer Mendelevitch, Vectara
AI Engineer Summit 2025 · 15:14
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- The Hidden Costs of Building Your Own RAG Stack — Ofer Mendelevitch, Vectara
Start here to understand how enterprise requirements shape the choice between building a RAG stack and using a managed service, including SaaS, VPC, and on-premises deployment options.
Ofer MendelevitchAI Engineer Summit 2025
- open-rag-eval: RAG Evaluation without "golden" answers.
Learn how UMBRELA and AutoNuggetizer assess retrieval and answers, and how the framework visualizes results and integrates with Vectara, LangChain, and LlamaIndex.
Ofer MendelevitchAI Engineer World's Fair 2025
- Enterprise Deep Research: The Next Killer App for Enterprise AI — Ofer Mendelevitch, Vectara
See how enterprise documentation can support automated RFP responses and onboarding guides, alongside deployment and security considerations.
Ofer MendelevitchAI Engineer Code 2025
Messages from the stage
The operational scope of a RAG stack
Ofer Mendelevitch contrasts self-managed RAG with Vectara's managed platform, including responsibilities for parsing, provenance, latency optimization, access controls, and operational staffing.
Separate evaluation signals
The open-rag-eval presentation distinguishes retrieval relevance, answer assessment, citation faithfulness, and hallucination detection, allowing different aspects of a RAG pipeline's output to be assessed.
Components of enterprise deep research
Ofer Mendelevitch describes combining agentic RAG, multimodal ingestion, hybrid retrieval, parallel agents, and HHEM hallucination detection for multi-step research.
Affiliations reflect each recorded session, not necessarily current employment.