Overview
Across 17 modules you build one agent: BrightCart's Returns Desk agent, which works a queue of return and cancellation requests unattended. It reads each request, looks up the order, applies policy, creates the return and refunds the money, within limits that code enforces and with a person approving anything expensive or irreversible. BrightCart is fictional; every lesson also opens with a real, cited case.
What you will learn
- 97 written parts, yours for good
- 33h 10m of reading, measured not estimated
- Written for the advanced level
- Every future revision included
- UPI, cards and netbanking
Prerequisites
- A Google account, to run code in Google Colab. Nothing to install.
- One free API key, from Groq or Google Gemini, to use in Colab. Or no key at all, if you run a model on your own computer with Ollama and work in a code editor.
- For a few lessons (container sandboxes, queues, deployment), your own computer. Those lessons say so.
Curriculum
18 sections · 97 parts · 33h 10m2h 11m free to read · 30h 59m with the course
01Course guide5 parts · 60 min
- Agentic AI Engineer: The Complete CoursePreview5 min
- How This Course WorksPreview6 min
- Setup: Colab, Your Model Provider, and Running LocallyPreview9 min
- The Project: BrightCart's Returns Desk AgentPreview10 min
- Shared Returns Desk DataPreview30 min
02Module 1: Prerequisites5 parts · 71 min
- Module 1: PrerequisitesPreview1 min
- Lesson 1: The plain engineering underneath every agentPreview20 min
- Lesson 2: Talking to a model: calls, tokens, schemas, retrievalPreview21 min
- Lab: The returns desk's first working slicePreview20 min
- Incident: Right numbers, wrong units, one lost spacecraftPreview9 min
03Module 2: Model Behaviour for Agents5 parts · 76 min
- Module 2: Model Behaviour for Agents1 min
- Lesson 1: What the model brings, and where it slips21 min
- Lesson 2: Steering the agent with a written prompt21 min
- Lab: A versioned prompt, tested on two models22 min
- Incident: A rule with no edges, applied everywhere11 min
04Module 3: Tool Design6 parts · 140 min
- Module 3: Tool Design1 min
- Lesson 1: How a model calls your tools, and how to help it29 min
- Lesson 2: Fewer, safer tools: designing the tool surface32 min
- Lesson 3: When tools fail: retries, limits and handing over30 min
- Lab: The eight tools, guarded and tested33 min
- Incident: An agent worked around a refusal15 min
05Module 4: Model Context Protocol6 parts · 127 min
- Module 4: Model Context Protocol1 min
- Lesson 1: One protocol for every tool: how MCP works29 min
- Lesson 2: Serving the returns desk tools over MCP34 min
- Lesson 3: When an MCP server works against you19 min
- Lab: The returns desk served and used over MCP37 min
- Incident: An email MCP server that copied everything7 min
06Module 5: The Agent Loop6 parts · 141 min
- Module 5: The Agent Loop1 min
- Lesson 1: The agent loop, written by hand30 min
- Lesson 2: Workflows, agents and the patterns between them29 min
- Lesson 3: Pausing, resuming and stopping a running agent28 min
- Lab: The agent loop moves into the package45 min
- Incident: A timeout retried 1,384 times in a minute8 min
07Module 6: Planning and Decomposition5 parts · 101 min
- Module 6: Planning and Decomposition1 min
- Lesson 1: Plans the agent can check, revise and finish27 min
- Lesson 2: Checking the work, not the agent's word26 min
- Lab: A checked plan for every request40 min
- Incident: Fixed, said an agent unable to test7 min
08Module 7: Context Engineering5 parts · 111 min
- Module 7: Context Engineering1 min
- Lesson 1: Why long runs go wrong: context is scarce32 min
- Lesson 2: Nine ways to keep the agent's context small30 min
- Lab: Budgets, compaction and notes for the shift37 min
- Incident: Long chat sessions confused Bing's model11 min
09Module 8: Memory and State5 parts · 110 min
- Module 8: Memory and State1 min
- Lesson 1: Keeping a run's state safe through a crash28 min
- Lesson 2: Memory that lasts between runs, and stays safe31 min
- Lab: Checkpoints, resumable runs and a careful memory41 min
- Incident: A bank's scheduled payments went out twice9 min
10Module 9: Retrieval for Agents5 parts · 109 min
- Module 9: Retrieval for Agents1 min
- Lesson 1: Search the agent drives, step by step26 min
- Lesson 2: Grep, SQL, files and past runs as tools31 min
- Lab: Search tools for the returns desk43 min
- Incident: References cited that nobody had read8 min
11Module 10: Environments and Execution6 parts · 129 min
- Module 10: Environments and Execution1 min
- Lesson 1: Letting the agent run code, inside walls21 min
- Lesson 2: Giving the agent a browser, safely23 min
- Lesson 3: Letting the agent touch real systems safely29 min
- Lab: A sandbox, a courier check and dry runs45 min
- Incident: A chart request that ran attacker code10 min
12Module 11: Multi-Agent Systems6 parts · 137 min
- Module 11: Multi-Agent Systems1 min
- Lesson 1: One agent first, and when to split it31 min
- Lesson 2: Orchestrators, handoffs, fan-out and swarms29 min
- Lesson 3: How agents hand work to each other safely27 min
- Lab: Decide whether to split, and prove it42 min
- Incident: Two pricing bots priced a book at $23.7 million7 min
13Module 12: Safety, Permissions and Oversight6 parts · 149 min
- Module 12: Safety, Permissions and Oversight1 min
- Lesson 1: The trust boundary: where outside words meet your tools29 min
- Lesson 2: Controls the model cannot talk its way past33 min
- Lesson 3: Seeing, explaining and reviewing what the agent did32 min
- Lab: One door, an audit trail and a kill switch46 min
- Incident: A trusted domain anyone could buy8 min
14Module 13: Evaluation5 parts · 117 min
- Module 13: Evaluation1 min
- Lesson 1: What to measure when an agent works alone29 min
- Lesson 2: Measuring an agent you cannot watch34 min
- Lab: An eval suite that gates every change38 min
- Incident: The grader read the answer, not the work15 min
15Module 14: Observability and Debugging5 parts · 111 min
- Module 14: Observability and Debugging1 min
- Lesson 1: Seeing inside every run, step by step31 min
- Lesson 2: Debugging a run: replay it, find the step, name the fault31 min
- Lab: Traces, replay and alerts for every run41 min
- Incident: An unattended loop found on the bill7 min
16Module 15: Runtime and Infrastructure6 parts · 138 min
- Module 15: Runtime and Infrastructure1 min
- Lesson 1: Running the agent as a job that survives a crash28 min
- Lesson 2: Serving the agent to many people at once28 min
- Lesson 3: Where the money and the seconds go33 min
- Lab: Run the desk as a budgeted background worker40 min
- Incident: Work that timed out faster than it finished8 min
17Module 16: Deployment and Operations5 parts · 104 min
- Module 16: Deployment and Operations1 min
- Lesson 1: Shipping the agent so you can undo it22 min
- Lesson 2: Keeping the agent healthy once it is live23 min
- Lab: Assemble the release, its drills and runbook50 min
- Incident: A blank policy field crashed every region8 min
18Module 17: Professional Practice5 parts · 59 min
- Module 17: Professional Practice1 min
- Lesson 1: Build less, promise less, check the claim15 min
- Lesson 2: Show the work: write-ups, READMEs, evals in public18 min
- Lab: Ship a repository you can show17 min
- Incident: A 98 percent claim without its setting8 min
Part 1 · freeAgentic AI Engineer: The Complete Course · 5 min
Systems that decide what to do next, act on the world, and keep going without you watching. Everything worth learning to build them, and to keep them from causing damage.
Across 17 modules you build one agent: BrightCart's Returns Desk agent, which works a queue of return and cancellation requests unattended. It reads each request, looks up the order, applies policy, creates the return and refunds the money, within limits that code enforces and with a person approving anything expensive or irreversible. BrightCart is fictional; every lesson also opens with a real, cited case.
Who this course is for
You know the AI engineer basics: Python, calling language models, prompting, structured output and retrieval. Module 1 is a quick check, not a first course.
Reviews
No reviews yet. The first comes from a reader who finishes the course.
to review this course once you have finished it.
Next to this one
Transformers: From Attention to Trained, Served and Interpreted Models111 parts · 43h 12m · ₹399Retrieval-Augmented Generation: The Complete Course95 parts · 38h 33m · ₹399Large Language Models: The Complete Course103 parts · 34h 48m · ₹399Forward Deployed Engineer: The Complete Course85 parts · 18h 15m · ₹399AI Engineering: The Complete Course105 parts · 23h 44m · ₹399
Large Language Models80 parts · 27h 0m · Free
Model Context Protocol83 parts · 14h 44m · ₹199
RAG82 parts · 13h 36m · ₹399
Python79 parts · 7h 28m · Free