Classes 10-12

Agentic AI Builder

“I can build AI systems that search, reason, remember and use tools.”

Your teen builds the most advanced kind of AI software there is, agents that plan, use tools and act, and deploys their own.

The capstoneBuild Your Own AI AgentPlans, picks a tool, runs it, checks the result, continues.

How the year runs

Length: 16 weeks
Live sessions: 32, with a mentor
Marked work: Practicals, with written feedback
Ends with: Build Your Own AI Agent
Parents see: Attendance, marks and feedback

Fees for the year₹19,999

Enrol for the year Meet a mentor first

An Alpha Plus plan for your child's class already opens this program's self-paced material. Enrolling adds the live part. What a year costs

The outcome

What your child walks out with

5 finished pieces of work, built one at a time and reviewed by a mentor. Not exercises: things that run, that they can open and show you.

01Research AgentPlans, picks a tool, runs it, checks the result, continues.

02Study AgentPlans, picks a tool, runs it, checks the result, continues.

03Document Intelligence AgentPlans, picks a tool, runs it, checks the result, continues.

04Multi-tool AssistantPlans, picks a tool, runs it, checks the result, continues.

05Personal AI AgentPlans, picks a tool, runs it, checks the result, continues.

Screens are drawn to show the shape of each build, not one student's work. Finished projects live in your child's own portfolio, which you can open from the parent dashboard.

Curriculum

What gets taught, module by module

12 modules, across 32 live sessions. Concepts are introduced when a project needs them.

LLM Engineering4 topics

LLM fundamentals
Tokens and context windows
System instructions
Structured outputs

Tool Calling3 topics

Functions and APIs
Tool use
External information

Embeddings & Search4 topics

Embeddings
Similarity
Vector databases
Semantic search

RAG4 topics

Documents and chunks
Retrieval and context
Citations
Basic evaluation

Memory3 topics

Conversation memory
User preferences
Short-term vs long-term

Agents3 topics

Chatbot vs agent
Goals, tools, planning
Actions, observations, feedback loop

Agent Workflows3 topics

State and branching
Loops and retries
Human approval

Multi-Agent Concepts1 topic

Practical, limited introduction

MCP Concepts1 topic

Modern tool interoperability, conceptually and practically

Safety4 topics

Prompt injection
Permissions and data handling
Guardrails
Human-in-the-loop

Evaluation3 topics

Correctness and reliability
Cost awareness
Latency awareness

Deployment1 topic

Ship the capstone agent

Week by week

What the time actually looks like

16 weeks, about 3.4 hours a week: 1.5 with a mentor, 0.9 of guided material and 1 of their own project time.

With a mentor: 24h; over 16 weeks
Guided material: 14.8h; over 16 weeks
Their own project time: 15.5h; over 16 weeks

Week by week, all 16 weeks

Week	Focus	Mentor	Guided	Own work
1	Readiness check, and what a model actually returns	90m	60m	40m
2	Calling a model like any other API	90m	60m	40m
3	Testing something that does not give the same answer twice	90m	60m	50m
4	Baselines: what would a simple approach get	90m	60m	50m
5	Preparing documents so they can be found	90m	60m	50m
6	Embeddings and retrieval	90m	60m	60m
7	Grounding, and abstaining when the source is silent	90m	60m	60m
8	Giving the model exactly one tool	90m	60m	60m
9	State: what the system remembers between steps	90m	60m	60m
10	Loops, and how to stop one	90m	60m	70m
11	Approvals: the actions a human must confirm	90m	60m	70m
12	Tracing: seeing what it did and why	90m	60m	70m
13	Failure tests, written to break it	90m	45m	70m
14	Cost caps and resource budgets	90m	45m	60m
15	Security and privacy review of your own system	90m	45m	60m
16	Capstone defence: an evaluated system and its failures	90m	30m	60m

What gets marked

3 assessed pieces of work. Everything else is practice.

A retrieval system that abstainsweek 7

Answers cite the passage they came from
It declines when the source does not contain the answer
A measured baseline exists to compare against

A traced, approval-gated tool loopweek 12

Every step is traceable after the fact
At least one action requires a human approval
The loop has a stopping condition that has been tested

The capstone, its evaluation and its defenceweek 16

Evaluation results are reported including the failures
There is a cost cap and evidence it holds
The learner defends a design decision under questioning

Optional extras: A second tool, added under the same approval rules, A monitoring dashboard for the trace data, A written incident report for a failure you caused deliberately

How it is marked

Automatic checks, AI suggestions and a mentor's decision are recorded separately, and a mark can be revised after feedback or appealed.

Functionality and accuracy35%
Reasoning30%
Testing and iteration20%
Communication and responsible use15%

How marking works is the same across every program.

Before week one

Checked before week 1. A learner who does not pass is offered Young AI Engineer first. Multi-agent work and MCP are extensions of this programme, not its entry point.

What is checked

Write Python that calls an API and handles an error response
Explain what a test proves and what it does not

Assumes your child can write Python and has worked with a model or an API. Multi-agent work is an extension of this programme, not its entry point.

Before you enrol

What your child needs

Said up front, because a laptop requirement discovered in week three is too late to discover it.

Prior knowledge

Comfortable with basic Python, or completion of Young AI Engineer.

Equipment

Laptop
8GB+ RAM recommended
Developer tooling (guided setup)
Stable internet

Tools they will use

PythonGitHubFastAPISupabaseLLM APIsVector databasesAgent frameworksCloud deployment

Every program runs the same way — learn it, build it, get it marked, with what you see as a parent the same throughout.

Where this goes

One year of nine, and it is built to continue

Alpha Euron runs from Class 4 to Class 12. Each program assumes what the one before it taught and hands over to the one after it, so a child who starts in Class 4 finishes school having built for nine years rather than having taken nine unrelated courses.

This is the last year on the ladder. From Class 8 a student can also carry their Alpha work across to the Euron platform, where the same account continues into adult courses and projects.

What they keep when it ends

A portfolio that is theirs

Every reviewed build stays in their account, openable and shareable, after the year ends.

A certificate that can be checked

Issued on completion with a number anyone can verify on this site, without a login.

Demo Day, in front of people

They present the capstone and answer questions on it. Most children have never done that.

The next year, already mapped

Each program ends where the next one starts, so nothing is repeated and nothing is skipped.

Questions

Before you enrol

Is this the right class for my child?

This program is designed for Classes 10-12. If your child is outside that range, or you are unsure, Find the Right Program recommends exactly one, and the free class confirms the fit.

What happens if they miss a class?

Lessons and guided practice stay available in the platform, and mentors help students catch up in the next session.

What equipment is needed?

A regular laptop with a browser is enough to start. Mentors help with any setup in class.

How are projects reviewed?

A mentor reviews every submitted project with marks and written feedback. Students can improve and resubmit.

Is this included in an Alpha Plus plan?

Partly. A plan for your child's class already opens this program's self-paced material. Enrolling adds the live part: scheduled sessions with a mentor, marked practicals and Demo Day.

Ready when your child is

The first class is free and includes a small hands-on project. Enrol whenever you are ready; the price is ₹19,999 for the whole program.

Enrol in Agentic AI Builder Book a free class