DataSense Guardian

• 

  Data Reliability Center — Real-time DataHub quality overview, schema intelligence, business metrics, and detected data risks.

• 

  DataSense Guardian — AI investigation connects data-quality issues to root causes, business impact, and recommended actions.

Inspiration

What is DataSense Guardian?

DataSense Guardian is an autonomous AI data reliability agent built with DataHub and MCP. It connects business datasets with metadata and data-quality signals, investigates problems automatically, and explains why those problems matter to the business.

Instead of only showing a data-quality score, DataSense Guardian answers the more important question: “What is wrong, why does it matter, and what should I do next?”

The Problem

Data teams often discover data problems only after inaccurate dashboards, financial reports, or business decisions have already been created.

For our demonstration, the orders dataset contains realistic data-quality risks:

• Missing values
• Duplicate order IDs
• A negative transaction amount
• Missing order_date metadata

These issues can lead to incomplete records, double-counting, unreliable financial calculations, and limited time-based analysis.

How We Built It

DataSense Guardian combines:

• DataHub for dataset metadata, schema information, and lineage
• DataHub MCP Server to connect the AI agent with DataHub context
• A local data profiler to calculate quality metrics and detect anomalies
• A backend API that coordinates investigation and business metrics
• A web dashboard that presents the investigation in a simple Data Reliability Center
• An AI investigation endpoint that converts technical data-quality findings into business-focused explanations

The agent investigates the dataset and produces evidence, impact, and recommended actions instead of simply reporting an error.

Example Investigation

For the orders dataset, DataSense Guardian identifies a 55/100 quality score and reports three major issues:

1. Missing values
2. Duplicate order IDs
3. One negative transaction amount

It also identifies that order_date is absent, limiting time-based financial analysis.

When asked why the dataset is dangerous for financial reporting, the Guardian connects these technical issues to business consequences such as potential double-counting, incomplete records, and unreliable financial analysis.

What We Learned

Building this project showed us that metadata becomes much more useful when an AI agent can reason over it together with actual data-quality signals.

We also learned how MCP can provide a structured interface between an AI agent and external data systems, allowing the agent to investigate instead of relying only on hard-coded responses.

Challenges

The biggest challenges were connecting the local application to DataHub, integrating the MCP server, keeping the frontend and backend investigation results synchronized, and making technical data-quality findings understandable to a business user.

We also focused on making the final experience simple: connect → investigate → understand the risk → take action.

Why It Matters

Data quality is not just a technical problem. A missing value or duplicate identifier can become a financial, operational, or decision-making problem.

DataSense Guardian turns those hidden risks into an actionable investigation so teams can understand the problem before it reaches downstream reporting.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for DataSense Guardian

Built With

• agents
• ai
• css
• data
• datahub
• docker
• engineering
• fastapi
• governance
• html
• javascript
• mcp
• observability
• python
• quality