Instruction file imported from drthgz/Enterprise_IT_Operations_Agent (
.github/instructions/project_instructions.instructions.md). Copyright stays with the author.
Requirement: Project Instructions
Agents Intensive - Capstone Project
Put your course learnings into action by building an agent that solves a real-world problem or improves everyday productivity.
Overview
This Capstone Project is part of the 5-Day AI Agents Intensive Course with Google (Nov 10 - 14, 2025) and is open to all participants. The project lets you apply what you’ve learned during the course by building AI agents in one of four tracks: Concierge Agents, Agents for Good, Enterprise Agents or Freestyle. Submissions are due by Dec 1, 2025, 11:59 AM PT, and top projects will win Kaggle swag, social media recognition on Kaggle’s platforms and all participants will receive a Kaggle badge and certificate.
Close
Dec 1, 2025 11:59 AM PT
Description
This capstone project is an opportunity for you to apply what you've learned during the course week and submit it for evaluation.
The high level process for a capstone project consists of:
Selecting a track (and only one track) from one of the four options. This will scope the problem/use case for your project.
Example: Concierge Agents
Formulating a problem and solution pitch to which you will build your agent. This will hone the objective of your project.
Example: Writing blogs is too manual and time intensive. I will be building an Automated Blog Writer Agent to scale my blog production and blog quality.
Developing and publishing the code for your agent publicly.
Example: Publishing the agent code to GitHub or Kaggle Notebooks.
Being able to articulate the value of the agent by preparing a writeup.
Example: This agent reduced my blog writing process by 10 hours per week
For bonus points, publishing a video related to the project.
Submitting a writeup to this competition no later than December 1, 2025 11:59AM Pacific Time. Check out "How Do I Make A Submission section" below.
Timeline
Submissions are due no later than December 1, 2025 11:59 AM Pacific Time
Winners will be announced before the end of December 2025.
Prizes
The top 3 teams in each track (including teams of 1) will receive Kaggle swag, and their projects will be featured on Kaggle’s social media platforms. All Capstone participants will receive a badge and a certificate of participation. Features to Include in Your Agent Submission
In your submission, you must demonstrate what you’ve learned in this course by applying at least three (3) of the key concepts listed below:
Multi-agent system, including any combination of:
Agent powered by an LLM
Parallel agents
Sequential agents
Loop agents
Tools, including:December
MCP
custom tools
built-in tools, such as Google Search or Code Execution
OpenAPI tools
Long-running operations (pause/resume agents)
Sessions & Memory
Sessions & state management (e.g. InMemorySessionService)
Long term memory (e.g. Memory Bank)
Context engineering (e.g. context compaction)
Observability: Logging, Tracing, Metrics
Agent evaluation
A2A Protocol
Agent deployment
Tracks and Awards
Agents for Good
Description
Agents that tackle problems in education, healthcare, or sustainability.
Track Awards
1st Place
Kaggle swag.
Non-monetary
2nd Place
Kaggle swag.
Non-monetary
3rd Place
Kaggle swag.
Non-monetary
Enterprise Agents
Description
Agents designed to improve business workflows, analyze data, or automate customer support.
Track Awards
1st Place
Kaggle swag.
Non-monetary
2nd Place
Kaggle swag.
Non-monetary
3rd Place
Kaggle swag.
Non-monetary
Concierge Agents
Description
Agents useful for individuals in their own lives, to automate meal planning, streamline shopping, plan travel, etc.
Track Awards
1st Place
Kaggle swag.
Non-monetary
2nd Place
Kaggle swag.
Non-monetary
3rd Place
Kaggle swag.
Non-monetary
Freestyle Track
Description
The open category for innovative agents that don't fit neatly into the other tracks. This is your space to experiment, explore, and build something truly unique or unclassifiable.
Track Awards
1st Place
Kaggle swag.
Non-monetary
2nd Place
Kaggle swag.
Non-monetary
3rd Place
Kaggle swag.
Non-monetary
Working as a Team
Your team can only make one submission for the capstone project.
You can submit as an individual participant - a team of one (1); or submit as a member of a team with the team having no more than four (4) members.
Each participant can only be in at most one team.
If you’d like to form a team or join an existing one, join Kaggle Discord and connect with other participants.
Evaluation
Your submission will be evaluated on three categories earning up to a maximum of 100 points.
Category 1: The Pitch (max 30 points)
Category 2: The Implementation (max 70 points)
Bonus: (max 20 points)
Note: Bonus points are added to the total of Category 1 and 2 points, up to a maximum of total 100 points.
Example scoring evaluation: Category 1 score of 30 points + Category 2 score of 60 points + Bonus points of 20 = 100 points. Criteria (points) Description Category 1: The Pitch (Problem, Solution, Value) (30 points total) This is where you'll be evaluated on the "why" and "what" of your project and how well you communicate your vision. Core Concept & Value (15 points) Your project's central idea, its relevance to the track for the submission; focused on innovation and value. The use of agents should be clear, meaningful and central to your solution. Writeup (15 points) How well your written submission articulates the problem you're solving, your solution, its architecture, and your project's journey. Category 2: The Implementation (Architecture, Code) (70 points total) This is where you'll be evaluated on the "how" of your project. This includes the quality of your code, technical design, and AI integration. Technical Implementation (50 points) In your submission, you must demonstrate what you’ve learned in this course by applying at least three (3) of the key concepts listed in the Features To Include In Your Agent Submission section.
For this criteria, we will assess the quality of your solution's architecture, and your code, and the meaningful use of agents in your solution.
Your code should contain comments pertinent to implementation, design and behaviors.
Participants are not required to deploy their agents to a live public endpoint for judging purposes; however, if you do deploy, please provide documentation to reproduce the deployment.
🚨REMINDER: DO NOT INCLUDE ANY API KEYS OR PASSWORDS IN YOUR CODE. Documentation (20 points) Your submission (when submitting via GitHub) should contain a README.md file explaining the problem, solution, architecture, instructions for setup, and relevant diagrams or images where appropriate.
If you are solely submitting a Kaggle notebook, please provide documentation directly inline via Markdown Cells of the notebook. Bonus points (Tooling, Model Use, Deployment, Video) 20 points total You can earn optional bonus points. Effective Use of Gemini (5 points) Use of Gemini to power your agent (or at least one sub-agent). Agent Deployment (5 points) If you either have code or otherwise show evidence (e.g. in your code or write up) of having deployed your agent using Agent Engine or a similar Cloud-based runtime (e.g. Cloud Run). YouTube Video Submission (10 points) Your video should include clarity, conciseness and quality of messaging. It should be under 3 min long. It should articulate:
Problem Statement: Describe the problem you're trying to solve, and why you think it's an important or interesting problem to solve.
Agents: Why agents? How can agents uniquely help solve that problem?
Architecture: Images and a description of the overall agent architecture.
Demo: demo of your solution, which can include images, an animation, or a video of the agent working.
The Build: How you created it, what tools or technologies you used. How do I make a Submission?
A submission consists of the following requirements via the Kaggle Competitions writeup process.
Watch this video to learn how to submit for this Capstone.
Note: You will not be able to submit your writeup unless you have satisfied these following requirements:
Title
Subtitle
Card and Thumbnail Image: select an image that can help identify your proposal quickly.
Submission Track
Media Gallery: [Optional] Add your YouTube video URL
Project Description (<1500 words)
Attachments: Add links to code, either as:
GitHub Repository: containing the code for your agent. This should be marked as publicly accessible in GitHub or applicable Git provider. OR
Kaggle Notebook: containing the code for your agent.
Resources
Development
ADK-Docs (link)
ADK-Python (link)
ADK-Go (link)
ADK-Java (link)
ADK-Sample Agents (link)
Agent Starter Pack (link)
A2A-Docs (link)
A2A-Python (link)
Google AI Studio (link)
Community
Reddit Agent Development Kit (link)
Documentation
Vertex AI Agent Engine Documentation (link)
Introduction to Agents Whitepaper (link)
F.A.Q.
Q: Can I make a submission multiple times?
No, you can only make ONE submission (so make sure it counts!).
Q: If I use ADK, can I choose another language (Java or Go), or do I need to stick with Python?
You can use ADK-Python, ADK-Go or ADK-Java.
Q: Do I need to deploy my agent to the cloud to qualify as a submission?
Deploying an agent is completely optional. However, if you share your code or describe it in your writeup, you will receive 5 bonus points.
Q: Can I make a submission using Colab or Kaggle Notebooks instead of Github?
Yes, you can either submit a Github repo, Kaggle Notebook or Colab notebook.
Q: How can I join or add people to my team?
To join an existing team, you will need to ask the team leader to add you to their team. To add people to your team, in the Team tab, add your teammates' Kaggle usernames. If you are looking to join a team with others, you can ask around on the Kaggle Discord.
Q: Can I still submit even though my code isn’t working or is not complete?
Absolutely. Your submission will still be evaluated on all criteria.
Q: Do I need to submit my write up and video in English?
Yes. In order to make the judging process fair and streamlined all submissions should be submitted in English.
My Project
Enterprise Agents
Description
Agents designed to improve business workflows, analyze data, or automate customer support. By analyzing emails, logs, and documents, the agent can identify common issues and provide instant responses to stakeholders and management. The agent can also generate reports summarizing key metrics and trends, helping the team make data-driven decisions to improve customer satisfaction and operational efficiency. Identify trends of outages and server resource usage to proactively plan for capacity and avoid downtime. The agents can take content in any format and curate it into a structured format for easy analysis and reporting. Also prepare dashboards that visualize key performance indicators (KPIs) and other metrics relevant to the enterprise. As well as suggest SLA, SLO and SLI metrics based on industry best practices and organizational goals. The agents will automatically scale itself and prepare or truncate tools as needed based on the complexity of the tasks at hand. +30 EP
Reference Notebooks
Use the notebooks in the notebooks folder as a reference implementation for building your own agents. These notebooks demonstrate various aspects of agent development, including multi-agent systems, tool integration, session management, and context engineering. You can adapt and extend these examples to suit your project's requirements. These notebooks were provided by Google as part of the Agents Intensive course.