AI Season curriculum

Key facts

The 12 sessions, week by week

1. W1 · Session 01 — AI Agents Foundations

   Understand how LLMs, tools, memory, and workflows become agents

   • Understand what AI agents are and why they matter
   • Learn how LLMs, tools, memory, and workflows connect
   • Build a simple agent that can follow instructions

2. W1 · Session 02 — Prompt Engineering & Vibe Coding

   Write prompts that hold up, and build with AI without shipping a mess

   • Write prompts that produce reliable, testable outputs
   • Use system prompts, examples, and constraints properly
   • Move idea to working app with AI-assisted coding

3. W2 · Session 03 — RAGs & Vector Databases

   Ground answers in your own data with retrieval and similarity search

   • Build a document Q&A agent over your own PDFs
   • Learn embeddings, chunking, and vector search
   • Judge retrieval quality instead of guessing at it

4. W2 · Session 04 — Design Patterns & Harness Engineering

   The proven shapes of agent systems, and the harness that runs them

   • Apply ReAct, planning, reflection, and routing patterns
   • Design the loop, context, and tool surface an agent runs in
   • Pick the right pattern instead of over-building

5. W3 · Session 05 — LangChain & LangGraph

   Build stateful, multi-step agent workflows with the standard frameworks

   • Use chains, prompts, tools, and retrievers
   • Build stateful agent workflows as graphs
   • Handle branching, retries, and multi-step logic

6. W3 · Session 06 — Mid Test & Revision

   Consolidate Sessions 1-5; this score counts toward your refund

   • Revise Sessions 1-5 with practical questions
   • Test understanding through hands-on tasks
   • Score toward the Top 3 refund challenge

7. W4 · Session 07 — Deployment

   Ship your agent to a public URL real users can actually reach

   • Take an agent from notebook to hosted service
   • Handle env config, secrets, and cold starts
   • Deploy your agent to a working public URL

8. W4 · Session 08 — AI Security & Guardrails

   Defend agents against prompt injection, leaks, and runaway cost

   • Spot prompt injection and unsafe tool calls
   • Add input, output, and permission guardrails
   • Cap cost, latency, and blast radius in production

9. W5 · Session 09 — MCP & Agent SDK

   Connect agents to real systems with the protocol the industry uses

   • Understand Model Context Protocol basics
   • Connect agents with external tools and systems
   • Use production-ready SDK patterns for cleaner agents

10. W5 · Session 10 — Conversational Voice Calling Agent

    Build an agent that answers the phone, listens, and talks back

    • Wire speech-to-text and text-to-speech into an agent
    • Manage turn-taking, interruptions, and latency
    • Ship a voice agent that handles a real call flow

11. W6 · Session 11 — Excalidraw Project

    Build the capstone: an agent that draws and reasons on a canvas

    • Turn a prompt into structured Excalidraw diagrams
    • Combine tool calling, state, and output validation
    • Ship one portfolio project end to end

12. W6 · Session 12 — Final Test & Refund Ranking

    Complete the final evaluation; top 3 scorers get a full refund

    • Complete the final practical evaluation
    • Combine all agent-building concepts learned
    • Top 3 scorers receive the full refund

How each session works

Each teaching session follows the same rhythm: a concept explained in Urdu with diagrams, a live coding walkthrough in English, and a task to finish before the next session. Questions are answered live and in the cohort WhatsApp group, and the recording is shared the same day. You write the code yourself — watching is not the goal.

What you will have built by week 6

• A first agent that follows instructions and calls tools (week 1)
• A document Q&A agent grounded in your own PDFs with RAG and a vector database (week 2)
• Stateful LangGraph workflows with branching and retries (week 3)
• An agent deployed to a public URL with secrets and configuration handled properly (week 4)
• Input, output and permission guardrails against prompt injection and runaway cost (week 4)
• Agents connected to external tools through MCP, using production agent SDK patterns (week 5)
• A conversational voice calling agent (week 5)
• An Excalidraw capstone — an agent that turns prompts into structured diagrams (week 6)

Tests and the top-3 refund

Session 6 is the Mid Test, covering sessions 1–5 with practical questions and hands-on tasks. Session 12 is the Final Test, combining everything. Your two scores are added together, and the three highest-scoring students in the cohort get their full fee back. Finishing the bootcamp earns a certificate of completion that anyone can verify online.

What you need before session 1

• A laptop that runs Python 3.10 or newer, and a stable internet connection
• Basic Python: functions, loops, lists and dictionaries, and calling an API
• A GitHub account to keep your projects
• Free-tier API keys from model providers (instructions are shared with the cohort)

Not there yet? The free how to learn AI roadmap covers the Python you need, and the open-source Cohort 01 code shows the level of the material.

How the curriculum was built

The syllabus follows how AI agents are actually built in 2026: start from the raw model and tool-calling loop, add retrieval, then frameworks for state and control, then the production concerns — deployment, security, evaluation and standard protocols like MCP. Cohort 01 taught this path live in July and August 2026, and its code is public on GitHub.

Frequently asked questions

How many sessions does AI Season have?

12 live sessions over 6 weeks: 10 teaching sessions plus the Mid Test and the Final Test.

Does the curriculum include LangChain and LangGraph?

Yes. Session 5 covers LangChain and LangGraph in depth, and later sessions build on them for deployment, guardrails and MCP-connected agents.

Is MCP (Model Context Protocol) covered?

Yes. Session 9 covers MCP and agent SDK patterns for connecting agents to real tools and systems. See the MCP guide.

Do I build real projects?

Yes — every teaching session ends with a runnable project, and week 6 is a capstone. You finish with a portfolio of working agents.

Can I see the course material before joining?

Yes. Cohort 01's code is open source — see the free AI agents course material.