Insightpro AI Education Analytics

The main idea behind InsightPro is that higher education institutions are data-rich but insight-poor. The goal of this project is to build a human-centric, ethical design framework that bridges the gap between complex machine learning and non-technical educators, turning intimidating spreadsheets into intuitive, plain-language visual stories.

My main focus was to design a low-barrier interface that takes complicated predictive data and translates it into real-world institutional decisions. I wanted to eliminate the slow, manual data management silos that academic leaders typically deal with, and give them a system where they can easily spot student risks early and take action before it’s too late.

I don't have immediate plans, but I’m definitely open to evolving the design further if the right opportunity arises.

One important realization came during our user research with university staff. We saw that many users were not struggling because they couldn't understand data, but because existing tools presented too much information at once. They had to switch between dashboards, spreadsheets, reports, and meetings just to answer simple questions.That changed how I approached the design. Instead of building another dashboard with more charts, I focused on reducing unnecessary cognitive effort. The interface gradually reveals information, combines visualizations with plain-language explanations, and helps users move from data to insight without feeling overwhelmed. That shift became one of the guiding principles of Insightpro.

Older generation analytics tools focused almost exclusively on historical, static descriptive data, which basically meant telling an institution what went wrong long after it was too late to act. Insightpro reframes this relationship by putting clear, predictive foresight directly into the hands of frontline staff like deans, program directors, and curriculum specialists. By utilizing an advanced machine learning backend to handle high-accuracy risk forecasting, the interface allows non-technical educators to immediately pinpoint exact subject areas that need immediate attention and intervention. It bridges the gap between raw data and the people who actually support the students.

Trust was a major consideration throughout the design process. I never wanted AI to replace human judgment or become the final decision maker. Instead, the system presents predictions together with supporting information and visual context, so educators can understand why a recommendation is being made. My goal was to design AI as a decision-support tool rather than an automated authority. Additionally, since the historical training data was heavily skewed toward average performers, we performed upsampling on minority student cohorts to prevent systemic algorithmic bias against struggling or top-performing students. In the UI itself, instead of issuing a rigid judgment, I designed node-specific annotations and the "Insights Inbox" so educators can layer their own real-world context directly onto data points, keeping final decisions strictly in human hands.