Free course
Free AI Agents Course Material
The code AI Season taught in Cohort 01 is free and open source on GitHub — runnable Python for calling LLMs, building a RAG pipeline, writing an agent harness from scratch, multimodal agents, LangChain, agent-to-agent (A2A) communication, a WhatsApp bot, and even training a small language model yourself. Here is what is inside and how to use it.
By Abdul Rahman Azam, founder of AI Season · Updated
Key facts
| Price | Free and open source |
|---|---|
| Where | github.com/AbdulRahmanAzam/AI-Season-Course-Material |
| From | AI Season Cohort 01 (July–August 2026) |
| Language | Python (plus JavaScript for a few demos); comments and READMEs in English |
| API keys | Most demos use free-tier APIs; several fall back to a mock mode with no key |
| Level | Beginner to intermediate — basic Python is enough to start |
What you get for free
Each folder is a self-contained lesson: numbered scripts you run in order, a README that explains what each file teaches, and a requirements file. The code is written for teaching — flat, readable scripts rather than a framework you have to understand first — so you can open any file and follow it top to bottom.
| Module | What you learn | Key tools |
|---|---|---|
| Session 1 — LLM API basics | Your first model call, streaming, chat history and parameters like temperature — in Python and JavaScript | Model provider SDKs |
| Session 3 — RAG from your own documents | Convert files, chunk, embed and store; retrieval methods and hybrid search; answer generation; failure experiments; a Streamlit chat app | LangChain, Chroma, local embeddings, Groq |
| Session 4 — Build an agent harness | The loop that turns an LLM into an agent: tools, memory, safety rules, a verifier and tests — built checkpoint by checkpoint | Plain Python, pytest |
| Session 5 — Multimodal agents | Vision-language models, audio, video understanding, OCR pipelines, document understanding, multimodal RAG and tool calling | Gemini API (free tier) |
| Session 6 — LangChain | Prompt templates, LCEL chains, runnables, structured output, memory, tools and agents, RAG basics, a leads agent | LangChain, FAISS, Groq |
| Session 8 — Production patterns | Honest RAG that refuses when unsure, and a self-correcting agent that fixes code until its tests pass | Python, pytest |
| A2A — agent-to-agent communication | Agent cards, agents that delegate to other agents, an orchestrator, the official A2A SDK and a two-agent debate | A2A protocol, Groq |
| ARGPT — build a language model from scratch | Tokenizer, transformer model, training on a rented GPU, generation and serving it behind your own URL — about 700 readable lines | PyTorch, Modal |
| WhatsApp bot | A real WhatsApp bot that answers messages with a LangChain chain | Node.js, Baileys, LangChain.js, Groq |
How to use the course material
- Install Python 3.10 or newer and Git.
- Clone the repository:
git clone https://github.com/AbdulRahmanAzam/AI-Season-Course-Material.git - Open one module folder and read its README first.
- Create a virtual environment and run
pip install -r requirements.txtinside that folder. - Copy
.env.exampleto.envand add a free API key where the README asks for one — or run without a key where a mock mode is available. - Run the numbered files in order, read each one, then change something and run it again.
Free API tiers have rate limits. If you see a "429 Too Many Requests" error, wait a minute and run the script again.
A suggested order if you are self-studying
- Session 1 — get comfortable calling a model from code.
- Session 3 — build RAG over your own PDFs; this is the most useful skill for real projects.
- Session 4 — build an agent loop by hand so frameworks stop being magic.
- Session 6 — learn LangChain now that you know what it wraps.
- Session 8 — make answers honest and agents self-correcting.
- Session 5 and A2A — extend agents to images, audio and documents, then to other agents.
- ARGPT — when you are curious what is inside a language model, build a small one.
Pair it with the free guides: how to learn AI, how to build an AI agent and the AI agents glossary.
What the free material does not include
Code is only part of learning. The repository does not include the live explanations, the Q&A, the tests, the feedback on your work or the certificate. Those are what the paid bootcamp adds — along with a fixed schedule and a cohort to learn with.
Want to learn it live?
AI Season teaches this material live in 12 sessions over 6 weeks, explained in Urdu with English code, for students in Pakistan, India and worldwide. Cohort 02 starts 1st January 2027 — PKR 3,000 early-bird, and the top three scorers get a full refund. Enrol here.
Frequently asked questions
Is the AI Season course material really free?
Yes. The Cohort 01 code is public on GitHub and free to use for learning. The live bootcamp — teaching, Q&A, tests and certificate — is the paid part.
Do I need a paid API key to run the code?
No. Most modules use providers with free tiers, and several scripts run in a mock mode without any key so you can follow the logic offline.
Is this a free AI course with a certificate?
The free material has no certificate. The certificate of completion is awarded to students who finish the live bootcamp, and it can be verified at aiseason.tech/verify.
Is there a free AI agents course in Urdu or Hindi?
The code and READMEs are in English. The live bootcamp explains the same material in Urdu, which Hindi speakers also follow easily.