DEV Community

Sandeep Nani
Sandeep Nani

Posted on

My Experience with Google-Kaggle AI agents Intensive Course

Recently I joined in the 5-Day AI Agents Intensive mostly out of curiosity. Everywhere I looked, people were talking about “Agents” but I felt like I only had a surface-level understanding of what that actually know it. I figured this course would either clear things up — or confuse me more. Luckily, it did the first one.

My Learning skills

1.What an Agent Actually Is

Before the course, I assumed agents were basically chatbots with extra steps.
Turns out there’s a whole structure behind them.

They’ve got:

  • something that plans,
  • tools they can use,
  • a bit of memory,
  • and a loop that lets them check their own work.

Once I saw it in action during the labs, it finally clicked. It’s not magic. It’s more like building a small system that happens to think with a model.

2. Multi-Agent Systems Are Logical

One of the sessions went into how multiple agents can work together.
I didn’t expect this part to be interesting,but it kind of changed how I think about complex tasks.Instead of forcing one model to juggle everything,you split the work:

  • one plans
  • one does the steps
  • one checks things

It felt oddly similar to how actual team projects work.

3. Tools Are the Real Power

This was my biggest “ohhh, I get it now” moment.

Agents get way more useful once they can:

  • run a bit of code,
  • call an API,
  • look something up,
  • read a file.

That’s when it stops feeling like a chatbot and starts feeling like software that can help you build things or automate tasks.

  • My Capstone Project

For my project, I built a simple goal-planning agent. You gave a task, and it would break it down into smaller steps and figure out what was needed.

Why I choose it:
I’m one of those people who writes lists for everything, so this feels natural.

What went well:
The planning part actually worked better than I expected. Separating the planner and executor helped keep it organized.

What didn’t:
The agent struggled with vague goals and also I had to change the tool definitions a lot because the agent kept calling them wrong.

What I learned:
Building an agent is half creativity and half debugging. Sometimes 80% debugging.

How My Understanding Changed

Before:

“Agents are just advanced chatbots.”

After:

“Agents are basically small systems powered by an AI model that uses tools, memory, and planning.”

It really changed how I think about AI projects. now I understand why people say agents are the next big step.

  • Final Thoughts

Overall, I’m really happy I learned the course.It gave me a clearer picture of how agents actually work and how to build something practical with them. I’m planning to keep experimenting, especially with multi-agent workflows, because that part was really excited.

Top comments (0)