I wrapped up an incredibly rewarding milestone🎯 and want to take a moment to reflect on my journey through the Google 5Day AgenticAI Course & Capstone Project.
🏃🏻♀️ Journey
Attending Google’s 5-Day Agentic AI Intensive Course in November was a great opportunity to learn and build end-to-end with Google ADK. Through a combination of 1) podcasts, 2) white papers, 3) live sessions, and 4) hands-on notebooks, I deepened my understanding of the fundamentals of building agentic applications the Google way.
The Capstone Competition Project offered the chance to apply what we learned by building with at least three of the following ADK-powered pillars:
▸ Agent Architectures — Parallel, Sequential, Loop, and Custom
▸ Agent Tooling & Integrations via MCP
▸ Context Engineering & Memory Management
▸ Agent Evaluation & Quality
▸ Agent Deployment
🗓️ Timeline
Week of Nov 10 — 👩🏻🏫 Absorbing content
Week of Nov 17 — 💆🏻♀️ Mind-mapping and ideation
Week of Nov 21 — 👩🏻💻 Bringing the pieces together and implementing
👩🏻🔬 Discovery
Before attending the intensive course, I had no clear idea of what I was going to build. The project emerged naturally—from the unknown—by blending my personal interest in the stock market with the tools and technologies explored during Agentic AI Week.
I decided to build a Stock Market Analysis Multi-Agent System to address a core need:
Stock traders need a single system that can gather market sentiment, run quantitative analysis, and clearly explain what it all means—instantly and coherently.
🚜🌾👩🏻🌾 What the End to End Looked Like
From logo design, ideation, problem definition, data availability checks, AI toolchain benchmarking (all happening in parallel!), to building the POC and delivering working examples—this project required full end-to-end product thinking and engineering skills.
I genuinely enjoyed the process, the problem-solving, and exploring what was possible with the tools at hand.
📦 Deliverable
I engineered and delivered a multi-agent stock market analysis system that provides actionable investment insights for S&P 100 traders.
The architecture leverages Google ADK as the core framework, using its multi-agent orchestration, function integrations, and MCP support for external tools. The system is powered by Gemini-2.5-Flash-Lite and Nemotron-9B-v2, enabling scalable analysis and reasoning.
🧗🏻♀️Challenges Along the Way
As with any build, the journey came with a few technical hurdles:
▸ Integrating the data science agent within Google ADK
▸ Loosing saved work and having to redo hours of progress
▸ Passing the Coding assistant tools' credit several times and nearing Kaggle GPU limits
✍️ Final Note
Thanks to Google and Kaggle teams for the learning opportunity.
Top comments (0)