AI Engineering Hub: 93+ Production-Ready LLM, RAG, and Agent Implementations
Explore patchy631/ai-engineering-hub, an open-source repository featuring 93+ hands-on AI engineering implementations, MCP workflows, and agent systems organize
The repository patchy631/ai-engineering-hub, created in October 2024 and updated in November 2026, has 39,811 GitHub stars and 6,447 forks. Its projects cover LLMs, Retrieval-Augmented Generation (RAG), Model Context Protocol (MCP), and multi-agent systems.
Image source: https://github.com/patchy631/ai-engineering-hub
Three-Tier Implementation Structure: Beginner, Intermediate, and Advanced
The repository organizes its self-reported catalog of 93+ projects into three distinct complexity tiers to support progressive skill adoption:
- Beginner (22 projects): Focuses on single-component workflows and foundational implementations. Highlights include local OCR using Llama, Gemma-3, and Qwen 2.5 models; lightweight chat interfaces built with DeepSeek-R1, Chainlit, and Streamlit; and fundamental document RAG pipelines powered by LlamaIndex and Ollama.
- Intermediate (48 projects): Covers multi-component systems and agentic workflows. Examples include a stock analyst using Microsoft AutoGen, content workflows using CrewAI, agentic RAG with document search and web fallback, real-time voice agents, and a RAG example reporting sub-15ms retrieval latency with Milvus and Groq.
- Advanced (23 projects): Covers complex systems, fine-tuning, and production deployments. Projects include fine-tuning DeepSeek with Unsloth and Ollama, building DeepSeek-R1-like reasoning models, implementing the Transformer architecture from scratch, and separate examples for MCP-powered multi-agent deep research and browser automation.
Model Context Protocol (MCP) and Workflow Observability
A dedicated segment of the repository centers on the Model Context Protocol (MCP), providing concrete integrations for modern development environments:
- MCP Clients and Integrations: Includes dedicated implementations such as web-search MCP connectors for Cursor, local MCP client setups with LlamaIndex, and an agentic RAG example using Firecrawl.
- Model Evaluation and Comparison: Includes an end-to-end RAG evaluation example using CometML Opik, alongside separate model and code-generation comparison projects.
Environment Requirements and Licensing Considerations
The repository is distributed under the MIT License.
The 93+ project count is the repository maintainer's own catalog. Practitioners should check each sub-project's setup instructions before running it.
Sources
- GitHub Repository: patchy631/ai-engineering-hub