RosaEduAgent - AI School Quality Assistant

Inspiration

What it does# About RosaEduAgent

What Inspired Me

I am a Quality Control Coordinator at an educational institution in Nigeria. Every week, I walk into classrooms, observe teachers, score their performance, and write reports. Then I watch those reports sit in a folder while the same problems repeat the next term.

The gap is not effort. It is not passion. It is infrastructure.

There is no system that takes what I observe in a classroom and automatically turns it into an action — a plan assigned to the right person, tracked to completion, visible to leadership. School quality in Nigeria collapses silently because the distance between seeing a problem and fixing it is too wide for any one person to close alone.

I built RosaEduAgent because I needed it. And because every school leader I know needs it too.

---

What I Learned

This project taught me that the hardest part of building with AI is not the technology it is the clarity of thought required to define the problem precisely enough for an agent to act on it.

I learned:

• How to structure teacher performance data in MongoDB so an AI agent can search and reason over it meaningfully
• How Google Cloud Agent Builder translates plain-English instructions into multi-step agent behaviour
• How Model Context Protocol (MCP) allows an AI agent to connect to a live database and retrieve real records not just generate text
• That the most powerful thing an agent can do is not answer a question, but complete a task

I also learned that building something from your own lived experience makes every technical decision easier. I was not guessing at what the data should look like. I have collected it by hand.

---

How I Built It

REduAgent was built using the following stack:

Layer	Technology
Reasoning & planning	Google Gemini 3
Agent orchestration	Google Cloud Agent Builder
Database	MongoDB Atlas
Agent-to-database connection	MongoDB MCP Server
Demo design	Canva

The architecture in three steps:

1. The data layer I created a MongoDB Atlas database called Rosaeduagent with a teachers collection. Each teacher record contains their name, subject, class level, competency scores across five domains (lesson delivery, classroom management, literacy instruction, assessment practice, and professional conduct), and a flag for whether an improvement plan is currently active.

2. The agent layer Inside Google Cloud Agent Builder, I wrote the agent's instructions in plain English — defining its role, its decision logic, and the actions it is permitted to take. The agent was given one tool: the MongoDB MCP server, which allows it to query and update teacher records in real time.

3. The interaction layer A school leader opens RosaEduAgent and types their challenge in natural language. The agent reasons over the input, queries the relevant teacher records, identifies the root cause, and generates a structured improvement plan with assigned owners, timelines, and follow-up checkpoints.

---

Challenges I Faced

1. Starting with no technical background

I am an educator, not a developer. Every tool in this stack was new to me. The biggest challenge was learning just enough to build without getting lost in complexity that was not necessary for this use case. I solved this by staying focused on one question at a time: what does this step need to produce?

2. Designing data that reflects reality

Creating the teacher database required me to think carefully about what information actually drives quality decisions in a school. Too much data and the agent loses focus. Too little and it cannot reason meaningfully. I drew on years of real observation forms, competency rubrics, and improvement frameworks I have built by hand.

3. Communicating a complex idea simply

The hardest sentence I wrote was the tagline: "It doesn't just answer questions. It acts." That single line took longer than most of the technical setup because it forced me to be precise about what makes an agent different from a chatbot, and why that difference matters for schools.

---

What's Next for RosaEduAgent

• Expand the database to support multi-school networks and local government education authorities
• Add a parent-facing literacy dashboard connected to student reading assessments
• Integrate with Nigeria's national curriculum standards for automatic alignment scoring
• Build a mobile-first interface for school leaders who manage quality from their phones

RosaEduAgent is not a hackathon project that ends here. It is the foundation of a school quality system that Nigeria's schools have needed for a long time.

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for RosaEduAgent - AI School Quality Assistant

Built With

• google-cloud-agent-builder
• google-gemini-3
• mongodb-atlas
• mongodb-mcp-server