AI Engineer

Runnable AI engineering modules: agents, MCP, memory, retrieval, voice, fine-tuning, MLOps, and security. Each folder is one module with its own README.

Why this repo?

Reading about AI engineering only goes so far. These modules run. You'll find:

• 19 modules in three tiers, from first scripts to full services
• Agents in LangGraph and the OpenAI Agents SDK, plus a 7-lesson MCP course
• Redis for memory and retrieval, and a Neo4j knowledge graph
• Deploys to AWS, GKE, and Kubernetes
• A live voice agent and an LLM-based code detector

Every module README lists what it is, how to run it, and what it needs.

Table of contents

• Getting started
• Projects by difficulty
  • Beginner (6)
  • Intermediate (8)
  • Advanced (5)
• Contributing
• License

Getting started

Each module is self-contained. Open its README, copy the keys it names into a .env, and run the commands.

1. New to this? Start with the Beginner modules, such as redis-basics and webhooks.
2. Build agents. Move to Intermediate with langgraph, openai-agents, and mcp-crash-course.
3. Go deeper. Try the Advanced modules: agentic RAG, Kafka pipelines, and Kubernetes.
4. Know the tags. Project is a runnable app. Tutorial is a set of small scripts. Course is an ordered series. Reference is material to read.

Most Python modules use uv. Python 3.13 or newer, and a current Node.js LTS for the TypeScript modules.

Projects by difficulty

🟢 Beginner

Single ideas and small deploys. Start here.

Agents

• OpenAI Agents SDK in TypeScript · Tutorial - Hello world, agent as tool, and dynamic instructions.

Memory and retrieval

• Redis basics · Tutorial - Redis strings and lists from Node.js and Python.

Events and integration

• Webhooks · Tutorial - An order notification system with FastAPI webhooks.

Security

• FastAPI authentication · Project - Argon2 password hashing and bearer-token login.

MLOps and cloud

• AWS EC2 with FastAPI · Tutorial - Deploy a FastAPI bookstore to an EC2 instance.

Prompting

• AI dev prompts · Reference - A step-by-step Cursor workflow from idea to PRD to tasks, and the GPT-4.1 prompting guide.

🟡 Intermediate

Agents, memory, voice, and CI/CD pipelines.

Agents

• LangGraph · Tutorial - Stateful agents as graphs: chatbots, ReAct, RAG, memory, workflows, and subgraphs.
• OpenAI Agents SDK · Tutorial - Guardrails, a manager agent, and router and triage patterns in Python.
• MCP crash course · Course - Seven lessons on the Model Context Protocol for Python developers.

Memory and retrieval

• Redis agent memory · Tutorial - Short-term and long-term memory for LangGraph agents on Redis.
• Knowledge graph · Tutorial - Build a graph from text with an LLM, and a Neo4j quickstart.

Voice

• ElevenLabs voice agent · Project - A live voice agent that looks up patient records and books appointments.

MLOps and cloud

• ML pipeline on GKE · Project - Train a model, serve it with Flask, and deploy to GKE with GitHub Actions.

Prompting

• Context engineering · Reference - A context engineering template for AI coding assistants, with PRP commands.

🔴 Advanced

Full services and production patterns.

Memory and retrieval

• Agentic RAG with Redis · Project - A RAG graph on a Redis vector store, with query rewriting and relevance grading.

Events and integration

• GitHub sync · Project - A GitHub dashboard with a React client, a FastAPI server, and a Kafka pipeline.

MLOps and cloud

• RAG on Kubernetes · Project - A RAG API on Kubernetes with Prometheus monitoring.

Security

• AI code detector · Project - An Express and TypeScript service that asks Claude if code looks AI-generated, with GitHub webhooks.

Fine-tuning

• Hugging Face fine-tuning · Tutorial - Fine-tune Qwen3-0.6B for support ticket routing and compare metrics before and after.

Contributing

Contributions are welcome. Read CONTRIBUTING.md first.

1. Open a Module proposal issue for a new module.
2. Fork the repository and create a branch.
3. Improve a module, or add a new top-level folder for one.
4. Copy the module README template. Add the module to the right tier above, with a type tag.
5. Open a pull request. The template lists the checks.

Report a leaked secret through private reporting. Everyone must follow the Code of Conduct.

License

MIT. See LICENSE.

Created by Kushal Banda.