A live LangGraph course for developers who want AI agents that are predictable, can pause for a human approval and can recover from errors. You model an agent as a graph of steps and state, then add tools, checkpoints, streaming and tests. Small batches, Hinglish classes, and a mentor reviews every assignment.
At a glance
Tools
Who it's for
Before you start
Please read: safe and age-appropriate use
Model providers have their own age rules and payment terms for API accounts. Students under 18 should set up accounts with a parent or guardian, and never share API keys.
Outcomes
Real skills you can show — every project goes on your GitHub.
Syllabus
Each module has a clear goal, live classes, a lab session and practice you push to GitHub. Your mentor reviews it and tells you exactly how to improve.
Module 1
Explain the problems with a free-running agent loop and how a graph helps.
PythonLangGraphLangChainGit
Practice: Draw the graph for a support agent on paper before writing any code
Module 2
Build a small working graph where each step reads and updates shared state.
LangGraphPydantic or typed dictionaries
Practice: A three-step writing pipeline: plan, draft, check
Module 3
Make the graph choose its next step and loop safely until the job is done.
LangGraphLangChain tools
Practice: A research agent that searches, reads, decides if it knows enough, and repeats within a limit
Module 4
Pause the agent before a risky step and continue only after a person approves, edits or rejects it.
LangGraph
Practice: A support agent that drafts a refund reply and waits for approval before it is sent
Module 5
Save progress so a run can resume, and show users what the agent is doing.
LangGraph checkpointersSQLite or PostgreSQL
Practice: Make your agent resumable and stream its progress to a simple web page
Module 6
Split a big job between specialised agents without creating chaos.
LangGraph
Practice: A supervisor agent that sends questions to a research worker and a writing worker
Module 7
Find why an agent failed, test it before release and stop it doing harm.
LangSmithpytest
Practice: Write ten scenario tests for your agent, including three that try to break it
Module 8
Ship an agent with approval steps behind an API and present it with honest notes on its limits.
FastAPIDocker (basics)Render or RailwayGitHub
Practice: Capstone: a research or support agent with approval steps, tests and a public repository
What a class looks like
So you know exactly what to expect before you join.
0–10
Recap, and a look at one failed run from last week's assignment.
10–40
Concept demo: the mentor draws the graph, then builds it live.
40–75
You build your version and the mentor helps with errors and design choices.
75–90
Read the trace together, discuss what could go wrong, and set the assignment.
Projects
Pick one, plan it like a real team with tickets, build it, and present it. It becomes the centrepiece of your portfolio.
Take a question, plan searches, gather sources, write a short report, and wait for a person to approve before it is saved or shared.
Read a message, look up order details with tools, draft a reply and ask a human to approve anything involving refunds.
A graph that drafts, checks facts and style, loops for fixes within a limit, and then asks an editor to sign off.
An agent that answers common IT questions and prepares a ticket, with approval before any account change.
Opening soon
This specialisation isn't scheduled yet. Join the waitlist and we'll tell you first when a batch opens. We open the courses with the most interest first.
Where it can lead
Roles this skill helps you towards
No course can guarantee a job. We help you build real skills, real projects and a portfolio you can show.
FAQ
The batch is not scheduled yet. Join the waitlist and we will tell you first when dates are fixed. The first batch is preceded by a free demo class.
It helps but is not compulsory. You should have called an LLM from Python and be comfortable with Python code. In the first module we recap what you need.
Yes. LangChain and LangGraph change quickly, so we review the syllabus before each batch and fix the exact library versions at that time. We teach the ideas and building blocks that stay the same, and the notes follow the current stable release.
Often you can, and we say so. A graph is worth it when you need approvals, resume after failure, clear limits or an audit trail. The course teaches how to decide.
Classes are in Hinglish. Notes, code and documentation are in English.
No. Nobody honest can. You will build a tested, deployed agent and a GitHub portfolio, and your mentor will give you feedback you can show to employers or clients.
Keep going
A live LangChain course for students who know Python and want to build apps on top of large language models. You connect a model to prompts, tools and your own documents, then trace, test and deploy a small app. Small batches, Hinglish classes, and a mentor reviews every assignment.
A live generative AI and LLM course for students who can already code in Python. You learn how large language models work, use LLM APIs and Hugging Face, and build document Q&A with RAG. Small batches, Hinglish classes and mentor review.
This backend development course teaches you to build the server side of real apps: REST APIs, databases, login and security, using Python, Django, Django REST Framework and PostgreSQL. Classes are live, in Hinglish, in a small batch, and a mentor reviews every assignment. A short overview module also shows how Node.js/Express and Flask do the same jobs.
This Python course takes you from your very first line of code to useful programs, games, scripts and data tools. Classes are live, in small batches of up to 30, taught in Hinglish, and a mentor reviews every assignment you push to GitHub. Python is taught inside our Starter and Builder programs.
Baking your code to give you a pie
Join a live demo class, meet the mentor and see how a CodeBakerz batch works. No payment, no pressure — decide after.
