patchy631/ai-engineering-hub

What this is

AI Engineering Hub collects hands-on tutorial projects for building with large language models: retrieval-augmented generation (RAG), agent workflows, MCP integrations, voice and multimodal apps, model comparison, fine-tuning, and production-oriented systems. The repository describes 93+ projects split by difficulty: 22 beginner, 48 intermediate, and 23 advanced project index.

Each project lives in its own directory with code and instructions, so the repo works less as an installable library and more as a catalog you browse and clone piece by piece. Beginners start with single-component apps such as OCR tools and simple RAG; intermediate work adds multi-agent flows and advanced RAG; advanced entries cover fine-tuning, MCP infrastructure, and production systems.

It suits developers learning to ship LLM applications, practitioners looking for reference implementations of RAG and agents, and researchers exploring current model stacks. Some entries depend on third-party services or local model runtimes, so individual projects vary in setup cost.

Get started

There is no repository-wide install command in the README; the intended path is to pick a project and follow its own directory instructions. A reasonable first pass:

1. Read the AI Engineering Roadmap if you are new to the field. It maps a path from Python basics to production AI.
2. Choose a beginner project matched to your goal. For a fully local RAG starting point, open Simple RAG Workflow, which combines LlamaIndex and Ollama; you need Python and a locally running Ollama instance with a model pulled before running it.
3. Move to an intermediate Agentic RAG example, which adds document search with a web fallback.
4. For deeper work, browse the advanced section for fine-tuning, MCP servers, or production pipelines.

The README does not list shared prerequisites, dependency versions, or a top-level environment file, so treat each project's own documentation as the source of truth for setup and expected output.

Constraints

• Projects are independent; there is no single package to install, and dependencies differ per directory.
• Many examples expect local model serving (Ollama or similar) or API keys for hosted providers such as Gemini, AssemblyAI, Firecrawl, or BrightData. Check the specific project before starting.
• GPU availability and model size requirements vary by project, particularly the vision, fine-tuning, and local chat examples.
• The repository is MIT licensed LICENSE, which covers the tutorial code but not necessarily the third-party models and services it calls.
• Newsletter subscription is optional and separate from using the code newsletter.

HotRepo's take

On the GitHub all-language daily list of November 1, 2026, the repository reached position 15 with 573 stars gained that day and a total of roughly 39.7k stars. That single-day record shows strong pull for a tutorial collection; no multi-day series is available here, so it supports a snapshot of attention rather than a trend judgment. The breadth of the catalog—local OCR, RAG stacks, MCP demos, and model comparisons—is what the ranking likely reflects: readers can find a runnable reference for nearly any common LLM application pattern without committing to one framework.