Nvidia partner bgr
NVIDIA Inception Program Member

NVIDIA
Integration Ready

Fabrix.ai agents run on the NVIDIA GPUs you already own. Connect them to NIM microservices or an open-source model stack, on-prem or in the cloud.

NVIDIA Inception Program — Fabrix.ai
  • NVIDIA GPU Acceleration
  • Nemotron & Open-Source LLMs
  • NIM Microservices
  • On-Prem · Private Cloud · NVIDIA Cloud

Integration Highlights

Fabrix.ai runs on standard CPU infrastructure. The NVIDIA integration is specifically about the inference layer: connecting Fabrix.ai agents to GPU-hosted LLM endpoints, wherever those endpoints live.

1

Your NVIDIA infrastructure hosts LLM endpoints

NVIDIA GPUs on-premises, in a private cloud, or via NVIDIA-hosted cloud endpoints serve language models for inference. This can be via NVIDIA NIM microservices (Path A) or an open-source stack (Path B).

Path A: NVIDIA NIM

With NVIDIA NIM Microservices

  • Fully managed
  • Enterprise SLA
  • Latency-critical
  • NVIDIA-first stack
  1. 2

    Deploy Nemotron, Llama, Mistral, or any supported model as a NIM container on your NVIDIA GPU servers. NIM handles drivers, optimization, and exposes a standard API endpoint.

  2. 3

    Configure the NIM endpoint in Fabrix.ai's Multi-LLM settings. Fabrix.ai agents immediately begin using that model for investigation, reasoning, and action across your entire IT estate.

Ideal for enterprises prioritising a fully managed, optimised inference stack with NVIDIA enterprise support and guaranteed SLAs.

Path B: Open-Source Stack

With Fabrix.ai-Assisted Install

  • Maximizing infrastructure ROI
  • Model flexibility
  • In-house infra expertise
  • Full stack control
  1. 2

    Fabrix.ai assists with installing CUDA libraries, NVIDIA drivers, and model-serving software (such as vLLM) directly on your GPU servers: on bare metal, private cloud, or hosted private cloud. No NIM license required.

  2. 3

    Open-source models, including GPT-OSS, Llama 3.x/4.x, Mistral, DeepSeek, Gemma, and others, are provisioned and served as endpoints. Fabrix.ai agents connect and operate exactly as they would with any other LLM.

Ideal for teams with in-house infrastructure expertise who want full control, model flexibility, and maximum return on their NVIDIA GPU investment.

Most enterprise teams use both

A hybrid approach lets you optimise cost and performance across different workload tiers.

NIM microservices for

  • High-priority production workloads
  • Enterprise-grade SLA requirements
  • Customer-facing agent interactions

Open-source stack for

  • Bulk inference at scale
  • Experimentation and model evaluation
  • Cost control across non-critical workloads
4

Fabrix.ai agents investigate, reason, and act, powered by your NVIDIA infrastructure

The Enterprise Knowledge Graph, agentic data federation, AgentOps observability, and all other Fabrix.ai capabilities operate normally, now with GPU-accelerated inference giving your agents faster, more capable reasoning at every step.

Typical Models on NVIDIA Infrastructure

Fabrix.ai connects to any model served via a compatible endpoint on your NVIDIA stack, NIM-managed or open-source. These are examples commonly deployed with Fabrix.ai for IT operations use cases.

NVIDIA Nemotron Family

  • Nemotron Nano
  • Nemotron Super
  • Nemotron Ultra
  • Llama Nemotron

Open-Source Models (via vLLM or NIM)

  • GPT-OSS
  • Llama 3.x / 4.x
  • Mistral 7B / 8x7B
  • DeepSeek
  • Gemma
  • Any OpenAI-compatible endpoint

Models are selected and validated per deployment based on your use case, hardware specifications, and inference performance requirements.

Real-World Deployment

Production-Grade Results on NVIDIA Hardware

A large Telco customer procured NVIDIA L40S GPUs on bare-metal servers and worked with Fabrix.ai to deploy an open-source LLM stack without any NIM licensing costs. Full data sovereignty, production-grade agent quality.

  • 100%On-premises data sovereignty; nothing leaves the network
  • ZeroAdditional NIM licensing costs; fully open-source stack
  • GPT-4Equivalent output quality for IT operations use cases

Deployment Stack

Tier-1 Telco · Production
Customer
Tier-1 Telecommunications Provider
GPU Hardware
NVIDIA L40S
Deployment
Bare-metal on-premises servers
Install Stack
CUDA + NVIDIA Drivers + vLLM
Model Serving
Open-source LLMs via vLLM
NIM Required
Not required
Data Residency
Fully on-premises

What You Get From This Integration

Whether you use NIM or the open-source path, the outcome is the same: agentic AI on your infrastructure, on your terms.

  • Complete Data Sovereignty

    Inference stays inside your environment. No data sent to external cloud LLM APIs: your IT telemetry, logs, and operational data never leave your infrastructure.

  • GPU-Accelerated Agent Reasoning

    NVIDIA GPU acceleration dramatically reduces inference latency, so agents investigate incidents, correlate events, and produce reasoning faster than CPU-only or shared cloud endpoints.

  • Maximize Your NVIDIA Investment

    GPU hardware you've already procured for training, simulation, or other AI workloads can now serve as the inference backbone for your entire IT operation agent fleet.

  • Full Model Flexibility

    Choose the model that fits your use case: Nemotron for agentic reasoning, Llama or Mistral for general intelligence, or GPT-OSS equivalents proven to deliver commercial-grade results at open-source cost.

  • Deploy Anywhere

    On-premises bare metal, private cloud, hosted private cloud, or NVIDIA-hosted public cloud endpoints. The integration works the same way regardless of where the GPUs live.

  • Full AgentOps Visibility

    Fabrix.ai's AI Observability layer traces every inference call, token cost, and model decision, giving you complete visibility into what your agents are doing on your NVIDIA infrastructure.

How Fabrix.ai Connects to Your NVIDIA Stack

Whether your endpoints run via NIM microservices or an open-source vLLM stack, Fabrix.ai agents connect, reason, and act across your entire IT environment.

The Fabrix.ai Agentic Platform connects to GPU-hosted LLM endpoints (on-prem, private cloud, or NVIDIA cloud) to power AI agents for IT operations.

Fabrix.ai NVIDIA Integration: GPU infrastructure connecting to Fabrix.ai agents for IT operations

Get Started

Start Your Agentic Journey
with NVIDIA Infrastructure

Talk to our team about your NVIDIA environment: GPU model, deployment type, and the models you want to use. We'll walk you through the integration options and validate the right configuration for your IT operations use case.

Request a Demo

On-prem · Private cloud · NVIDIA-hosted cloud · NIM or open-source · Any compatible LLM endpoint