Welcome to Build Your Own Super Agents — a fast, fun, no-fluff adventure into agentic AI. In these notebooks, you’ll train agents, evolve them, and even make them battle (yes, there are Pokémon). You’ll start with a tiny single-step agent and end up commanding full swarms that can plan, retrieve, reason, and act. Whether you’re a researcher, engineer, or curious tinkerer, this course gives you everything you need to level up from simple prompts to real super-agent systems. 🚀🌟

🎯 Course Highlights

• Fully hands-on agentic notebooks for every module — clone locally or run instantly in your preferred Jupyter environment.
• Practical agent templates, evaluators, and tool integrations so you can build real agents without drowning in boilerplate orchestration code.
• Clear theory, references, and best-practice design patterns interwoven with examples to make every agentic concept intuitive and reproducible.
• Production-ready mindset throughout — covering retrieval, graph reasoning, multi-agent coordination, reflection loops, model routing/selection, cost-latency tradeoffs, and cloud-scale deployment.

📚 Course Structure

Each module is a self-contained notebook filled with explanations, demos, and practical agentic workflows. You can browse them on GitHub Pages, clone them locally, or run them interactively in your preferred notebook environment.

Module	Topic	Notebook
01	A Very Simple Agent	01-simple-agent.ipynb
02	Moving to Agentic Frameworks	02-framework-pydantic-ai.ipynb
03	Prompt Engineering & Reflection Loops	03-prompting-engg.ipynb
04	Retrieval-Augmented Generation (RAG)	04-rag.ipynb
05	Knowledge Graphs & GraphRAG	05-graphrag.ipynb
06	Agentic Evaluation & Continuous Confidence	06-evaluation.ipynb
07	Multi-Agent Workflows & Agent Swarms	07-multi-agent-workflows.ipynb
08	Model Selection for Agents	08-model-selection.ipynb
09	Model Placement (Edge, Cloud, Hybrid)	09-model-placement.ipynb
10	Appendix: End-to-End Experimentation	10-appendix-experimentation.ipynb
11	Appendix: Designing Scalable Agentic Systems (AWS)	11-scalable-agentic-systems.ipynb
12	Bonus: Agentic Reinforcement Learning	12-bonus-agentic-rl.ipynb

🧠 What You'll Learn

• How to design an agent’s full loop — perception, retrieval, reasoning, planning, and action.
• How to use modern agentic frameworks to add tools, memory, structured outputs, and multi-step workflows.
• Techniques for building powerful retrieval systems (RAG), knowledge graphs, and GraphRAG pipelines.
• How to evaluate agents with confidence metrics, behavioural tests, and multi-agent benchmarking.
• How to build, coordinate, and optimize multi-agent swarms that collaborate or compete.
• Practical orchestration: model selection, model placement, cost–latency tradeoffs, and routing strategies.

⚙️ Quick Start

Option A: Launch in Lightning Studio (no setup!)

1. Click the Open in Studio badge above.
2. Authenticate with Lightning (or create a free account).
3. Add API Keys in .env file. Follow .env.example and this.
4. Explore the notebooks in a fully provisioned environment.

Option B: Run Locally

1. Clone the repository

   git clone https://github.com/shreshthtuli/build-your-own-super-agents.git
   cd build-your-own-super-agents

2. Install dependencies (recommended: Python 3.10+)

   pip install uv
   uv sync

3. Add API Keys in .env file. Follow .env.example.
4. Install docker by following steps here.
5. (Optional) Create LogFire account here and log in using

   logfire auth

6. Launch Jupyter

   jupyter lab

7. Open any notebook to start experimenting.

🧭 Suggested Learning Path

1. Foundations (Modules 01–03) – Build your first agent, understand agentic loops, and layer in prompting + reflection.
2. Retrieval & Knowledge (Modules 04–05) – Add RAG, build knowledge graphs, and integrate GraphRAG into agent workflows.
3. Evaluation & Behaviour (Module 06) – Learn how to measure agent performance, confidence, robustness, and reasoning quality.
4. Multi-Agent Systems (Module 07) – Coordinate multiple agents, share context, resolve conflicts, and design agent swarms.
5. Orchestration (Modules 08–09) – Master model selection, routing, placement, and the cost–latency–quality trade-offs that power real systems.
6. Capstone – Combine everything to build, evaluate, and deploy a production-ready super-agent tailored to your real use case.

Every notebook stands alone, but following this sequence gives you the smoothest progression from simple agents to fully orchestrated super-agents.

🧑‍💻 Who am I?

I’m Shreshth Tuli, an AI researcher, engineer and educator specializing in advanced ML systems, optimisation and real-world deployment. I build and teach systems that go beyond isolated models into full pipelines and platforms. Done this course? Feel free to connect on LinkedIn @shreshth-tuli or GitHub @shreshthtuli.

More about me here.

🤝 Contributions

Contributions, ideas, and bug reports are always appreciated! To contribute:

1. Fork the repository and create a feature branch.
2. Open a pull request explaining what you changed and why.
3. Reference any affected notebooks or modules, and include examples or screenshots if helpful.

Check the issue tracker for beginner-friendly tasks or start a discussion if you’d like to propose a new agent module, framework, or workflow.

📄 License

This project is released under the Apache 2.0 License. You're welcome to use the material in your own projects, demos, workshops, or derivative courses — just keep the attribution intact.

Don’t just study agents—forge your own and unleash it. 🚀