Introduction:-
I am Dhruv Mishra a Computer Science with IOT student at SRM University, Chennai, India. And in November I participated in the 5-Day Intensive Course on AI Agents by Google and Kaggle.
This was my first time learning about AI agents. I heard about the course on Instagram and was profound to join the course and learn how to make AI agents.
My most of the exposure to Ai was limited from IOT and basic ML concepts and using LLMs as prompting tools for researching. I had first heard of "AI agent" from a concept by ChatGPT where they introduced an agent which could open few websites or extensions and help the user digitally to provide them help like purchasing something from store help the user to how to proceed with it. I was very excited about this and this course help me get into that mindset once more and a more practical way to think about AI systems.
My Initial Understanding and Doubts:-
Before I was introduced with this course i thought AI agents were just:
- Better prompting tools
- More advanced or complex chatbots
- Or advanced versions of LLM
Honestly speaking during the first sessions many concepts like planning, tool usage, understanding code felt a bit abstract to me. I understood the words but didn't understand how they all functioned together. So searching watching videos, live streams and even gemini helped me clear understanding and kept me engaged.
One Thing That Changed My Perspective:-
The biggest realisation was that AI agents are systems, not prompts.
This helped me to think that how to ask model the right question is not the correct thinking it is how should the system decide what action to perform next.
An agent:
- Has a goal
- It breaks that goal into steps
- Chooses tools when needed
- Evaluates its own output
This felt very similar to how we design software systems, which made it easier for me to relate as CSE student
Learning About Planning and Tool Usage:-
One thing that i would like to appreciate was the importance on planning before acting upon it. Tool usage was another important concept. Instead of expecting the model to know everything agents are designed to:
- Fetch data
- Execute code
- Call required APIs
This way felt more realistic and closer to real world applications are built
Practice and Capstone Experience:-
The practical practice of the course helped me theory with reality. When i tried building a simple agent workflow for capstone, I realized that its not about making agents work once it's about making stops at right time and work consistently. I did struggled a bit with deciding when the agent should stop and defining clear steps. But these helped me more than just reading concepts.
For the capstone, I applied the concepts to build a simple agent-based solution.
The hardest part was controlling the agent’s behavior and making sure it didn’t repeat steps unnecessarily.
This taught me:
- The importance of clear stopping conditions
- Why evaluation steps matter
- How small design decisions affect agent reliability
Even though it was basic, it helped solidify my understanding.
Multi-Agent Systems:-
The idea of multiple agents working together was new to me, I didn't even realised it could have existed. Instead of one agent doing everything, the course showed how different agents can take on specific roles and collaborate and work together.
This reminded me of different types of operating system services that can be applied when an AI is introduced to Operating System. Thinking of more possibilities of agents.
How This Course Helped Me as a CSE IOT Student:-
As a computer science student, this course helped me see AI agents as a software engineering challenge, not just an AI or ML problem.
It connected well with concepts like:
- Modular design
- System architecture
- Debugging workflows
- Iterative improvement
It also made advanced AI topics feel more approachable, even for someone experiencing agentic AI for the first time.
What I Want to Explore Next:-
After completing the course, I want to:
Obvious that my dream agent project is to build something like what ChatGPT initiated an AI agent that would do or help you to do all over the Operating System.
But for now as my skill level:
- Build small, simple agents on my own
- Experiment with agent frameworks
- Understand common failure cases better
I don’t feel like an expert after five days of course, but I now have a clear starting point and a much better mental model.
Final Thoughts:-
For someone learning about AI agents for the first time, this course was a great balance of theory and practice.
It didn’t assume deep prior experience, but it also didn’t oversimplify the ideas.
The Google and Kaggle AI Agents Intensive Course helped me understand where AI is heading, and how I, as a student, can start preparing for it.
Top comments (0)