AI Fluency & Building · Agentic AI Developer
Go deeper. Real-world tasks that need more research, planning and a polished result.
Task 1 of 3
Write a cited brief on any company.
What to do
Build a research agent that takes a business name, searches public sources, and writes a 1-page brief: what they sell, who their customers are, their online presence, and 3 growth ideas, with citations. It must flag anything it could not verify.
You’ll make
Done looks like
An agent that produces a cited 1-page brief and flags unverified points.
Task 2 of 3
Step limits, allowed tools, cost caps and approvals.
What to do
Add guardrails to your agent: a step limit, an allowed-tool list, a cost cap, and a human approval step before it posts, sends or publishes anything.
You’ll make
Done looks like
Your agent with all 4 guardrails working and a short demo of each.
Task 3 of 3
Measure accuracy, cost and time on 10 questions.
What to do
Create a 10-question test set with expected answers, and measure your agent's accuracy, cost and time per task.
You’ll make
Done looks like
A 10-question test set with expected answers and measured results.
Learn before you build
Free course
A free hands-on agents course with a certificate.
Free course
Free courses on building stateful, multi-step agents.
Free course
Dozens of free 1–2 hour courses on agents, RAG and evaluation.
Code repo
Free notebooks on the API, tool use and evaluations.
Article guide
Read this first: when to use agents, and the patterns that work.
Article guide
A lightweight open-source framework for multi-agent workflows.