Solution Blueprints Overview#
Solution Blueprints are reference applications built with AIMs. Solution Blueprints offer an easy way to explore AIMs in the context of a complete microservice solution, such as document summarization, RAG chatbots, AI coding assistants, and agentic workflows. For developers, Solution Blueprints act as starting points and example implementations, making it fast and easy to solve real-world needs with ROCm software.
The industry standard orchestration system for microservices such as AIMs is Kubernetes. Solution Blueprints are packaged as helm charts, a templating software for Kubernetes manifests. Solution Blueprint charts are ready to deploy out-of-the-box to an AMD Enterprise AI Platform cluster, while Helm templating allows them to be customized for specific needs.
AI models are powerful tools, but they are rarely useful on their own. They need an interface to real-world, an integration into a solution. Yet building applications with AI models at their core is not trivial. Solution Blueprints come with documentation and architecture diagrams that make it clear how the pieces fit together.
Catalog#
The Solution Blueprint catalog ranges from simple demos and basic interfaces to fully-fledged agentic platforms and development environments. The catalog is constantly expanded to showcase the cutting edge innovations and to cover the established AI use cases.
Blueprint |
Description |
AIMs |
Resources |
|---|---|---|---|
Decoupled RAG blueprint with an MCP Knowledge Server and an agentic UI client. |
GPT OSS 20B, Embedding, ChromaDB |
2 GPU, 13 CPU, 272Gi RAM |
|
AI-powered UI testing with an LLM agent that interprets specs, automates browser tests, and generates CI-ready test code. |
GPT OSS 120B |
1 GPU, 8 CPU, 72Gi RAM |
|
Multi-agent translation workflow where LLM agents critique and refine outputs to improve translation quality. |
Llama 3.3 70B |
1 GPU, 5 CPU, 68Gi RAM |
|
No-code visual platform to design, test, and deploy agentic AI workflows with language models and tools. |
Llama 3.3 70B |
1 GPU, 6 CPU, 68Gi RAM |
|
Agentic documentation assistant that analyzes repositories and generates structured software architecture and component docs. |
Llama 3.3 70B |
1 GPU, 4 CPU, 64Gi RAM |
|
Browser-based AI pair programmer with a local LLM for code completion, fixes, and interactive coding support. |
Qwen 3 32B, Qwen 2 5 Coder 7B |
2 GPU, 8 CPU, 192Gi RAM |
|
Multimodal summarization service for text, PDF, DOCX, audio, and video content. |
Llama 3.3 70B, Whisper |
1 GPU, 4 CPU, 64Gi RAM |
|
Financial analysis assistant combining stock data, technical indicators, and LLM reasoning for market insights. |
Llama 70B |
1 GPU, 5 CPU, 68Gi RAM |
|
Sandbox chat interface to test prompts, evaluate outputs, and understand LLM behavior before production use. |
Llama 3.1 8B |
1 GPU, 5 CPU, 68Gi RAM |
|
Prompt router that classifies requests and sends them to the best LLM endpoint using configurable routing rules. |
Llama 3.1 8B, Llama 3.3 70B, Embedding |
1 GPU, 4 CPU, 192Gi RAM |
|
MRI analysis tool with a Gradio interface and LLM-generated reports. |
GPT OSS 20B |
1 GPU, 4 CPU, 68Gi RAM |
|
End-to-end pipeline that turns PDF documents into podcast-style audio using agentic orchestration, LLMs, and TTS. |
Llama 3.3 70B |
2 GPU, 16 CPU, 104Gi RAM |
|
Multi-agent preventive healthcare outreach system that identifies screening candidates and drafts personalized emails. |
Llama 3.3 70B |
1 GPU, 5 CPU, 68Gi RAM |
|
Multi-stage LLM engine that performs web research and produces structured, citation-backed technical reports. |
Llama 3.3 70B |
1 GPU, 2 CPU, 68Gi RAM |
|
RAG application for chatting with documents using ChromaDB retrieval and LLM-based answers. |
Llama 3.3 70B, Embedding, ChromaDB |
2 GPU, 11 CPU, 268Gi RAM |
|
Real-time telecom voice assistant for customer support, account actions, and ticketing workflows. |
Livekit, Qwen Asr, Openai GPT OSS 120B, Qwen Tts, ChromaDB, Embedding |
4 GPU, 29 CPU, 350Gi RAM |
Source Code#
Since Solution Blueprints are intended for developer reference, it can be useful to browse the source files. All AMD-built Solution Blueprint helm charts can be found in a public Git repository here.
Additionally, the helm charts can be pulled from the registry to browse their source files. See instructions here.