The latest articles on Forem by Md Mahadi Hasan (@mahadiosb). https://forem.com/mahadiosb https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3650407%2Fe06a739a-ae62-4787-b544-ae768800532a.jpg https://forem.com/mahadiosb en Md Mahadi Hasan Sun, 07 Dec 2025 16:18:38 +0000 https://forem.com/mahadiosb/my-5-day-journey-into-ai-agents-2al3 https://forem.com/mahadiosb/my-5-day-journey-into-ai-agents-2al3 <h2> AI Agents Intensive Course – Google x Kaggle </h2> <h2> Introduction </h2> <p>I joined the <strong><a href="https://www.kaggle.com/learn-guide/5-day-agents" rel="noopener noreferrer">5-Day AI Agents Intensive Course with Google and Kaggle</a></strong> to understand how modern AI agents work and how they can be used to solve real-world problems.<br><br> Before this course, I only used simple chatbots. After this course, I learned how to build <strong>real multi-agent systems with tools, memory, and workflows</strong>.</p> <p>This is my learning journey and project reflection.</p> <h2> 🗓️ Day 1 – From Prompts to Real Agents </h2> <p><strong>What I learned:</strong></p> <ul> <li>Difference between a chatbot and an <strong>AI agent</strong> </li> <li>How agents can take actions instead of only replying</li> <li>Basic agent architectures</li> </ul> <p><strong>My reflection:</strong><br><br> This was the biggest mindset shift for me. I learned that agents are not just “smart replies” — they are systems that can <strong>think, decide and act</strong>.</p> <h2> 🛠️ Day 2 – Agent Tools &amp; Interoperability </h2> <p><strong>What I learned:</strong></p> <ul> <li>How agents use <strong>tools (functions / APIs)</strong> </li> <li>How tools help agents interact with external systems</li> <li>Best practices for safe tool usage</li> </ul> <p><strong>My application:</strong><br><br> I created <strong>appointment tools</strong> that allow my agent to:</p> <ul> <li>Create new hospital appointments</li> <li>Fetch existing appointment details from a CSV database</li> </ul> <p>This made my agent feel like a real working system.</p> <h2> 🧠 Day 3 – Sessions &amp; Memory (Context Engineering) </h2> <p><strong>What I learned:</strong></p> <ul> <li>How agents remember conversations</li> <li>Session handling and memory storage</li> <li>Why memory makes agents feel more human</li> </ul> <p><strong>My implementation:</strong><br><br> I connected my main agent with:</p> <ul> <li><code>InMemorySessionService</code></li> <li><code>InMemoryMemoryService</code></li> </ul> <p>Now my hospital agent can <strong>remember user context during a session</strong>.</p> <h2> ✅ Day 4 – Agent Quality &amp; Observability </h2> <p><strong>What I learned:</strong></p> <ul> <li>How to measure agent quality</li> <li>Observability: tracking what agents think and do</li> <li>Why evaluation matters in production agents</li> </ul> <p><strong>My reflection:</strong><br><br> This day helped me understand that building an agent is not only about features — it's about <strong>trust, reliability, and safety</strong>.</p> <h2> 🚀 Day 5 – Multi-Agent System (Agent2Agent) </h2> <p><strong>What I learned:</strong></p> <ul> <li>How agents communicate with other agents</li> <li>Agent2Agent (A2A) architecture</li> <li>How to deploy agents</li> </ul> <p><strong>My capstone implementation:</strong></p> <p>I built a <strong>multi-agent hospital system</strong>:</p> <ul> <li>A <strong>Remote Appointment Agent</strong> (handles CSV data)</li> <li>A <strong>Hospital Orchestrator Agent</strong> (talks to users and calls the remote agent)</li> </ul> <p>This design makes the system scalable and modular.</p> <h2> 🏥 My Capstone Project – <a href="https://www.kaggle.com/code/mahadiosb/hospital-appointment-orchestrator-agent" rel="noopener noreferrer">Hospital Appointment AI Agent</a> </h2> <h3> Project Idea </h3> <p>Many hospitals still use manual appointment systems.<br><br> I built an <strong>AI agent system</strong> that can:</p> <ul> <li>Book appointments</li> <li>Assign rooms automatically by time</li> <li>Fetch appointment details instantly</li> </ul> <h3> Architecture </h3> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fvay0s781gm9a3f4wttxp.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fvay0s781gm9a3f4wttxp.png" alt="Flow" width="800" height="404"></a></p> <p>User → Orchestrator Agent → Remote Appointment Agent → CSV Database<br> This architecture helped me understand <strong>real-world agent system design</strong>.</p> <h3> 🎓 Capstone Project Demo </h3> <p>🎥 <strong>Demo Video:</strong> <a href="https://www.youtube.com/watch?v=F0c6Xfe7FCE" rel="noopener noreferrer">https://www.youtube.com/watch?v=F0c6Xfe7FCE</a></p> <h2> 🎯 Key Learnings </h2> <ul> <li>Agents are <strong>action-driven</strong>, not only chat-driven</li> <li>Tools turn agents into real applications</li> <li>Memory and sessions create personalized experiences</li> <li>Multi-agent systems are powerful and scalable</li> </ul> <h2> 🙏 Gratitude </h2> <p>I sincerely thank <strong>Google and Kaggle</strong> for organizing this amazing <strong>5-Day AI Agents Intensive Course</strong>.<br><br> It gave me hands-on experience and confidence to build real AI agent systems.</p> <p>This course changed how I see AI — from simple chatbots to intelligent, autonomous systems.</p> <h2> 🔮 If I Had More Time </h2> <p>If I had more time, I would:</p> <ul> <li>Add voice support to the agent</li> <li>Build a web UI dashboard</li> <li>Connect it with real hospital APIs</li> <li>Add authentication and security layers</li> </ul> <h2> Final Thoughts </h2> <p>This course was not just learning — it was <strong>building, experimenting, and growing</strong>.<br><br> I am excited to continue my journey in <strong>Agentic AI</strong>.</p> <p>Thank you! 🚀</p> googleaichallenge ai kaggle gemini