Relevance Labs Windows GenAI Workspace - GPU-Ready AMI

Stop spending days configuring GPU drivers, resolving CUDA version conflicts, and wiring together AI frameworks. This Windows Server 2022 AMI delivers a fully preconfigured generative AI research environment with 30+ tools ready to use from first login via Amazon NICE DCV remote desktop.

Unlike Linux-based Deep Learning AMIs, this is the only Windows-native workspace that bundles LLM fine-tuning tools (LoRA/PEFT), a pre-wired RAG chatbot pipeline (Bedrock + OpenSearch), medical imaging frameworks (MONAI), and multimodal AI pipelines (Whisper/CLIP/LLaVA) - all accessible through a full GUI desktop with Visual Studio 2022, RStudio, PyCharm CE, and JupyterLab side by side.

• ML engineers and data scientists who need a Windows-based GPU workstation in the cloud
• Research teams in healthcare leveraging MONAI for medical imaging alongside LLM experimentation
• Enterprise developers building RAG applications with AWS Bedrock and OpenSearch
• Teams prototyping GenAI applications who want pre-integrated sample workflows rather than starting from scratch

Get productive immediately with included working examples:

• RAG Chatbot - A complete retrieval-augmented generation pipeline using AWS Bedrock and OpenSearch, ready to customize
• Multimodal Pipelines - Pre-configured Whisper (speech), CLIP (vision-language), and LLaVA workflows
• Medical Imaging - MONAI framework configured for healthcare AI research

The AMI ships with Windows Firewall enabled, remote desktop access encrypted via Amazon NICE DCV protocol, and only the DCV port (8443) open by default. Enterprise and healthcare teams should review their own compliance requirements and apply additional hardening as needed for their regulatory context.

Recommended Instance Types:

• g4dn.xlarge - Entry-level evaluation (16 GB GPU memory)
• g5.xlarge - Standard GenAI development (24 GB GPU memory)
• p3.2xlarge - Large model training and inference (16 GB GPU memory)

Prerequisites:

• Security group allowing inbound TCP port 8443 (for NICE DCV remote desktop)
• IAM role with permissions for Bedrock, SageMaker, and OpenSearch (if using AWS-native integrations)
• Bedrock model access must be enabled separately in your AWS account
• Minimum 100 GB EBS volume recommended

Getting Started:

1. Launch the AMI on a supported GPU instance from AWS Marketplace
2. Wait for instance status checks to pass
3. Connect via NICE DCV client using your instance's public IP on port 8443
4. Log in with your Windows credentials (Administrator account)
5. Open JupyterLab or any preinstalled IDE and run the included sample notebooks

GenAI and LLM Frameworks: Hugging Face Transformers, LangChain, LlamaIndex, FAISS, LoRA/PEFT

ML and Data Science: PyTorch, TensorFlow, scikit-learn, PySpark, Dask, Vowpal Wabbit, MONAI

Development Environments: Visual Studio 2022, VS Code, PyCharm CE, JupyterLab, RStudio Desktop and Server

AWS Integrations: Bedrock SDK, SageMaker SDK, OpenSearch connectors, AWS CLI

Productivity and Infrastructure: Docker, Docker Compose, Anaconda, LibreOffice, Google Chrome, Git, 7-Zip

Languages: Python 3.x, R

Want to see the workspace in action before committing? Contact Relevance Labs to schedule a guided pilot session where the team walks you through the RAG pipeline, MONAI setup, and multimodal workflows on a live instance tailored to your use case.

• Supported on GPU instance families: g4dn, g5, p3, p4d
• NVIDIA drivers and CUDA Toolkit are pre-installed - no manual GPU configuration needed
• Model weights for LLaVA, Whisper, and CLIP may require separate download depending on model size
• Windows Firewall is enabled with DCV port configured
• This AMI does not include model weights for large foundation models - use Bedrock or download from Hugging Face Hub

The workspace comes pre-configured with SDKs and connectors for AWS Bedrock (foundation model access), SageMaker (model training and deployment), and OpenSearch (vector search for RAG applications). All integrations use standard AWS credential chains - configure your IAM role and start building immediately.