I am not a software engineer. I work for a company that develops real estate, but I've been studying data science, machine learning, and AI for a while now.
This year, I decided to move from practice into production. But where I had first hoped to build an agent, it turned out to be more of an app.
The Pivot: From Wind to Markets
The first iteration didn’t start with real estate; it started with the wind. I went to Google Cloud Next 2025 earlier this year, sketching out ideas for a sailing app to help newbies get "out of irons."
I realized the same logic used to navigate oceans could also navigate residential and commercial markets. By September, the sailing app had evolved into an intelligent platform designed to cut real estate feasibility analysis from weeks to minutes.
Building the Architecture
To get there, I had to move beyond what I had learned about deploying a live streaming app at sea to building a Hybrid Agent.
My learning path accelerated through several key events:
- Attended Google's AI Agents - Live followed by DevFest (both in Atlanta), where the idea to construct an intelligent architecture began to take shape.
- That momentum carried into November with a one-day ADK seminar.
- Participated in the hands-on lab: "Securing the Chain of Trust for Multi-Agent Systems on Cloud Run."
All of these events were part of the "Legos" for exploring what an agentic system could really be.
The Catalyst: Kaggle Intensive
December brought validation. I passed the Generative AI Leader Certification, helping to confirm the need for real estate intelligence was sound.
But the week-long Kaggle Intensive on AI Agents is what actually connected the dots. All of them. From multi-agent architectures to SecOps to agent limitations. Each day, I left with at least one—if not many—new ideas to explore. The content was always grounded in reality but presented amazing, unthought-of possibilities as well.
Just like university, I left with plenty of material to review after the online sessions. The daily guides I now use as reference materials to ensure my agents are reproducible and scalable. These 5 days helped reinforce everything I’ve been learning all year.
Moving to Production
I’ve picked up that learning quickly is an asset when getting into the thick of things. The program was well-structured to enable that style of learning while leaving enough wiggle room to explore other concepts.
My project, structured with the Google ADK, is currently going through refinements to make it suitable for enterprise users. Integrating graph database technology as an agent tool has allowed me to move from simple data retrieval to "Graph-Native Intelligence".
I am very grateful to have participated in the 5-Day Intensive Agents Course and look forward to future programs. Enough thinking—time to build.
Top comments (0)