NVIDIA
NVIDIA is the dominant GPU manufacturer powering AI training and inference infrastructure. In this wiki it also appears as an enterprise AI platform vendor: AI-Q, NeMo Agent Toolkit, Nemotron/NIM, and Dell AI Factory reference architectures extend NVIDIA’s role from chips into deployable agent infrastructure.
AI-Q Deep Research Blueprint (2026-05)
NVIDIA AI-Q is an open-source deep research blueprint that can be exposed to agent harnesses such as Claude Code, Codex, LangChain Deep Agents, and OpenCode as a portable skill. The pattern is to let the harness delegate research to a local or hosted AI-Q server and receive a cited report, while AI-Q owns retrieval, planning, synthesis, authentication, and evaluation.
Key points from NVIDIA AI-Q Deep Research Skill:
- Skill as pipeline adapter:
SKILL.mdplus a helper script submits research jobs to AI-Q instead of forcing every harness to rebuild retrieval and citation logic. - Enterprise data boundary: AI-Q can run where sensitive data lives; the harness receives the cited answer, not unrestricted raw document access.
- MCP integration: authenticated MCP servers are connected as NeMo Agent Toolkit function groups, with patterns for unauthenticated, service-account, and signed-in user token flows.
- Deployment flexibility: Docker Compose and Helm support laptop, cloud/on-prem Kubernetes, and air-gapped deployments; open Nemotron models can run through NVIDIA NIM when self-hosting is required.
- Auditability: cited reports plus OpenTelemetry traces make retrieval and synthesis inspectable for compliance teams.