NVIDIA Agent Knowledge Foundation

NVIDIA's models and libraries for grounding agents in enterprise data: Nemotron retrievers, Kumo models and Auto-Ontology.

NVIDIA · Data, training, and fine-tuning platforms (datasets, labeling, RAG/vector DBs, fine-tuning-as-a-service) · tool

Published in Feed

Overview

NVIDIA introduced the Agent Knowledge Foundation in the GTC Berlin keynote on October 21, 2026, a set of models and libraries that turn company data into something agents can search, interpret and use for predictions without moving data or changing governance. It combines CUDA-X libraries (cuDF, cuGraph, cuML, cuOpt, cuVS), a new Auto-Ontology library on GitHub that maps raw data to business concepts such as customers, products and orders, Nemotron Retriever open models, and a new Kumo open-model family (Tabular, Relational, Time Series, Forecast). It sits inside the open-source NVIDIA Agent Toolkit alongside OpenShell and NemoClaw blueprints. NVIDIA's supply chain organization uses it through Palantir Foundry, Databricks pairs Kumo Anomaly with Genie Ontology, and Snowflake customers can use Kumo Engine and Kumo Relational in Snowflake Notebooks. Reported Nemotron Retriever results: SAP's fine-tuned reranker ran nearly 3x faster and ServiceNow cut reranking latency by nearly a quarter.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Nemotron Retriever models for document parsing, embedding and reranking
  • Kumo open models: Tabular, Relational, Time Series and Forecast, for structured and time-series predictions
  • Auto-Ontology library maps raw company data to business concepts and their relationships
  • CUDA-X libraries: cuDF, cuGraph, cuML, cuOpt, cuVS
  • Part of the open-source NVIDIA Agent Toolkit with OpenShell runtime and NemoClaw blueprints
  • Demonstrated with Lila Sciences, Siemens Erlangen and Dassault Systemes

Use cases

  • Retrieval over enterprise documents for agents
  • Predictions from tables such as customer and payment data
  • Grounding engineering and science agents in proprietary data

Pricing

Unknown; NVIDIA's posts give no pricing, release dates for each component or detailed license terms.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
21 Oct 2026
Record updated
27 Oct 2026
Record last checked
27 Oct 2026

Explore the Feed

Other entries filed under data, training, and fine-tuning platforms (datasets, labeling, rag/vector dbs, fine-tuning-as-a-service); inclusion does not imply equivalent features.

  • DigitalOcean Agent Data Services

    DigitalOcean's managed data layer for agents: ingestion flows, a catalog with vectors, and SQL or vector query, in private preview.

  • pplx-embed-v2-late

    Perplexity's late-interaction multimodal embedding models (0.6B and 9B) with 128-dim per-token vectors, open on Hugging Face.

  • EmbeddingGemma 2

    Open-weight 740M-parameter multimodal embedding model built on Gemma 4 that encodes text, images, audio, and video for on-device search.

Browse the Feed