Skip to content
50% offFounding price until 1 DecemberSee prices
Square1 Ai
Get started
CoursesLLM Agent Architect

LLM Agent Architect

beginner-advanced

The builder track: ship production LLM agents — tool use, memory, MCP, and the harness engineering that makes them reliable.

14 modules109 lessons10 projects
Take assessment

Never written code? You can still start here.

Module 0 opens with 18 lessons of programming from nothing: your first program, variables, loops, functions, reading errors, the terminal, Git and arrays, before any of the LLM Agent Architect material. No prior coding assumed.

What you'll learn

~165 h guided · about 33 weeks at 5 h/week · +149 h on projects

  • Explain why ChatGPT and Claude generate answers rather than looking them up like a search engine
  • Describe a program as ordinary text saved in a file and read by Python one line at a time
  • Understand the difference between deterministic and stochastic testing for LLM agents
  • Design goal-decomposition pipelines that break objectives into executable sub-tasks
  • Explain the M×N integration problem and how MCP standardizes the interface to reduce it to M+N
  • Define an agent harness and identify its core components: the agentic loop, tools, context management, guardrails, and termination conditions
  • Explain what an augmented LLM in a loop is and enumerate the six capabilities frameworks add on top of it (orchestration, state/checkpointing, memory, tool wiring, multi-agent, observability)
  • Explain why the screenshot-reason-action loop emits one small action at a time, re-observes after every action, and runs under a step budget

Curriculum

0/109 lessons
1

AI Foundations — optional basics (free)

Week 0 · 20 lessons

0/20
1

Beyond the chatbox: what AI assistants really are

2

How the model "thinks" — and why it changes how you ask

3

The capability map: what AI does brilliantly

4

The limits: hallucination, freshness, and what never to trust it with

5

Your first real win: turn a 30-minute task into 3

6

Context is everything: give it what it needs to know

7

Give it a role and a goal

8

Constraints and format: define what "good" looks like

9

Show, don't tell: examples and templates

10

Iterate like a pro: steer, don't restart

11

Documents and reports: summarise, extract, and draft better

12

Spreadsheets and data: Excel formulas, analysis, and charts

13

Presentations: from blank page to a solid deck

14

Long, messy input: transcripts, threads, notes, and screenshots

15

ChatGPT, Claude, Copilot, Gemini: which to reach for, and when

16

Make smarter decisions: AI as your thinking partner

17

Chaining steps: multi-part tasks done reliably

18

Your prompt library: reusable templates that save hours

19

Staying safe: privacy, accuracy, and company data

20

Capstone: do a real task end-to-end (graded) → your certificate

2

Module 0 — Foundations: LLMs, Prompts, Tools & the Agent Toolkit

Week 0 · 23 lessons

0/23
1

What a program is and how to run your first one

2

Storing values in variables, and why text and numbers behave differently

3

Making decisions with if, else and comparisons

4

Storing many values in a list

5

Looking things up by name with dictionaries

6

Doing something to every item with loops

7

Writing your own functions

8

Reading errors without panic

9

Finding the mistake in your own code

10

Saving your code in a file and running it

11

Importing code other people have written

12

Giving each project its own environment

13

Talking to your computer with the terminal

14

Saving your work with Git

15

Putting your work on GitHub

16

What an array is, and why a list is not enough

17

Picking out pieces: indexing and slicing, including two dimensions

18

Maths on a whole array at once, and where you go next

19

How LLMs Work: Tokens, Context & Sampling

20

Prompts & the Messages API

21

Tools, JSON Schemas & Function Calling

22

Python & APIs for Agents

23

Git & API-Key Hygiene for Agent Developers

3

Agent Foundations

Week 1 · 5 lessons

0/5
1

What Are AI Agents?

2

Agent vs Chatbot

3

The Observe-Think-Act Loop

4

Agent Design Patterns

5

Agent Evaluation

4

Tool Use

Week 2 · 5 lessons

0/5
1

Function Calling Basics

2

Defining Tool Schemas

3

Handling Tool Results

4

Multi-Tool Agents

5

Tool Error Handling

5

Agent Loops

Week 3 · 5 lessons

0/5
1

ReAct Pattern

2

Plan-and-Execute

3

Iterative Refinement

4

Self-Correction

5

Loop Termination

6

Memory Systems

Week 4 · 5 lessons

0/5
1

Short-Term Memory

2

Long-Term Memory

3

Episodic Memory

4

Semantic Memory with Vectors

5

Memory Retrieval Strategies

7

Multi-Agent Systems

Week 5 · 5 lessons

0/5
1

Orchestrator Pattern

2

Specialist Agents

3

Agent Communication

4

Debate & Consensus

5

Hierarchical Agents

8

RAG for Agents

Week 6 · 5 lessons

0/5
1

Agent Knowledge Bases

2

Dynamic Retrieval

3

Context Window Management

4

Grounded Responses

5

Knowledge Updates

9

Production Agents

Week 7 · 5 lessons

0/5
1

Human-in-the-Loop

2

Safety Guardrails

3

Cost Control

4

Observability

5

Agent Testing

10

Advanced Patterns

Week 8 · 5 lessons

0/5
1

Autonomous Workflows

2

Agent Chains

3

Self-Improving Agents

4

Agent Benchmarking

5

Deploying Agent Systems

11

MCP & Agent Protocols

Week 9 · 7 lessons

0/7
1

Why MCP: The M×N Integration Problem

2

MCP Architecture: Hosts, Clients, Servers & the Lifecycle

3

Tools: Exposing Capabilities to the Model

4

Resources & Prompts: Context and Templates

5

Sampling, Roots & Elicitation: Inverting Control

6

MCP Security: Trust, Tool Poisoning & OAuth 2.1

7

A2A & the Protocol Landscape: Agent-to-Agent + Observability

12

Agent Harness Engineering

Week 10 · 7 lessons

0/7
1

The Agent Harness & the Workbench Loop

2

Scope Contracts & Repo Memory

3

Verification Gates: Making the Loop Self-Correct

4

Reviewer Agents & Multi-Session Handoff

5

Eval-Driven Agent Development

6

Agent Benchmarks: SWE-bench, GAIA, WebArena, OSWorld

7

Capstone — Build & Measure a Coding-Agent Harness

13

Modern Agent Frameworks

Week 11 · 6 lessons

0/6
1

The Framework Landscape & Build-vs-Buy

2

The Claude Agent SDK

3

The OpenAI Agents SDK

4

LangGraph: Stateful Graph Orchestration

5

CrewAI & AutoGen: Multi-Agent Patterns

6

Choosing & Migrating Between Frameworks

14

Computer-Use & Browser Agents

Week 12 · 6 lessons

0/6
1

Agents That See Screens: The Computer-Use Model

2

Browser Automation Foundations: DOM, Selectors and Waiting

3

The Observe-Think-Act Loop

4

Reliability: Verification, Checkpoints and Human Gates

5

Evaluating GUI Agents

6

Production Patterns for Web Agents

Advanced tier

Advanced LLM Agent Architect

6 senior modules · 30lessons · capstone & certificate. Upgrade your plan to unlock the advanced tier.

Advanced RAG & Knowledge Systems

5 lessons

1

Agentic RAG Architectures

2

Graph RAG & Knowledge Graphs

3

Multi-Modal Retrieval

4

Adaptive Chunking & Indexing

5

Production RAG Pipelines

Agent Frameworks & Infrastructure

5 lessons

1

LangGraph Deep Dive

2

CrewAI & Role-Based Agents

3

AutoGen & Conversational Agents

4

Custom Framework Design

5

Framework Selection & Migration

Enterprise Agent Security

5 lessons

1

Prompt Injection Attack Taxonomy

2

Input Sanitization & Defense Layers

3

Data Exfiltration Prevention

4

Sandboxing & Execution Isolation

5

Compliance & Audit Trails

Advanced Multi-Agent Orchestration

5 lessons

1

Swarm Intelligence Patterns

2

Agent Markets & Auction Mechanisms

3

Emergent Coordination & Self-Organization

4

Scaling Multi-Agent Systems

5

Cross-Organization Agent Federations

Agent Reliability Engineering

5 lessons

1

Evaluation Frameworks at Scale

2

A/B Testing Agent Systems

3

Observability & Distributed Tracing

4

Incident Response for Agent Failures

5

Continuous Improvement Pipelines

Capstone & Research Frontiers

5 lessons

1

World Models & Reasoning Agents

2

Tool-Making Agents

3

Agent-Computer Interfaces

4

Ethics & Societal Impact of Autonomous Agents

5

Capstone Project Workshop

What you'll build

10 projects · ~149 h · each graded against a rubric

  1. Tool-Using Chatbot

    beginner · ~8 h

    PythonAnthropic SDKFunction Calling

  2. Code Executor Agent

    beginner · ~10 h

    PythonAnthropic SDKSubprocess

  3. Web Research Agent

    beginner · ~10 h

    PythonAnthropic SDKTavily

  4. Multi-Step Reasoning Agent

    intermediate · ~14 h

    PythonAnthropic SDKReAct

  5. Agent with Memory

    intermediate · ~14 h

    PythonAnthropic SDKVector DB

  6. Multi-Agent Debate

    intermediate · ~16 h

    PythonAnthropic SDK

  7. Autonomous Task Planner

    intermediate · ~16 h

    PythonAnthropic SDKDAG

  8. Human-in-the-Loop Agent

    advanced · ~18 h

    PythonAnthropic SDKFlask

  9. Agent Evaluation Framework

    advanced · ~18 h

    PythonAnthropic SDKPytest

  10. Production Multi-Agent System

    advanced · ~25 h

    PythonAnthropic SDKRedisFastAPI

What we track

Everything below is graded by Nova and kept on your record: the thing an employer can open, run and verify.

Get started

Take the free assessment to see your skill level and get a personalised learning plan.

Take assessment

Featured projects

Tool-Using Chatbot

PythonAnthropic SDK

Code Executor Agent

PythonAnthropic SDK

Web Research Agent

PythonAnthropic SDK

About

14 modules, 109 lessons
~165 hours of guided learning
10 portfolio projects (~149h)
AI-graded exercises

LLM Agent Architect

20 questions · ~30 minutes

Take the assessment →

The people behind this track

Nova grades every submission; these are the humans who wrote the brief and stand behind it.

Nikhil De Silva

Founder, Square1 Ai

Built Square1 Ai to close the gap between knowing and doing: every lesson ends in work that gets graded, and every student has someone in their corner.