Designing a 0→1 AI product that turns fragmented academic data into one system.
Handed Off
Dec 2025
Industry
EdTech / Higher Education, B2B2C
Role
Lead Product Designer
Team
8 Student Designers, Salesforce Experience Design Team
TL;DR
Students had to piece together academic decisions across disconnected university systems, with no personalized guidance connecting courses, schedules and long-term goals.
As one of two lead designers, I designed Galileo, an AI companion integrated into IUB’s iGPS platform. I defined the information architecture, designed core planning experiences and established explainable AI patterns for trustworthy recommendations.
Before and After
What actually changed for students, before Galileo and after.
Feature 1: In-depth course review generated from university portals and peer reviews.
All in one system
In-depth course and professor reviews by peers, a 10,000 feet overview of their academic history and connections to majors and careers, under one tool.
Easy sync between planner and enrollment
The plugin carries a student's plan directly into iGPS, so registering picked up where planning left off instead of starting over.
Platform 1: iGPS portal of IUB for enrollment into classes
Limited and scattered information
4 disconnected systems
Degree audits, course catalogs, enrollment portals, and peer advice lived in separate places, leaving students to manually connect courses to majors and careers.
Enrolling and planning had separate flows
Planning and registering lived in separate systems, so a finalized semester plan had to be manually re-entered or re-searched.
Early choices locked in long-term consequences
Workload balance and major eligibility down the line got shaped by decisions made before a student had enough context to make them well.
Scattered inputs wore down confidence and time
Manually synthesizing academic data from all these sources increased cognitive load and left students less sure of their own choices.
Design Approach
From interviews and a scope matrix to one academic planning tool
01
Screening survey + semi-structured interviews
Of twenty-four survey responses, 91% named academics and coursework as their top priority, which is what pointed the interviews at academic planning. With over 100 synthesized insights, we had a lot to go over.
02
Importance x Opportunity matrix for scope cut
Academic challenges scored as the strongest actionable opportunity. Mental health scored just as high on importance, but we ruled it out since an AI system shouldn't be the one replacing a licensed professional.
03
Undeclared students are most vulnerable
They were the extreme case: making independent academic decisions for the first time, with no fixed major to anchor their course choices against.
04
6 rounds of testing with students
Early versions tried to cover academic, social and career support at once. Every round pushed the team to cut further until only academic planning was left standing.
Final Solution
Academic Trajectory: naming how the AI knows, and where it might be wrong
Transparent, explainable recommendations
I showed what data informed each suggestion and refined the recommendation logic to balance both completed coursework and student interest.
The four lenses of academic progress
Lets a student view the same course history by year, by potential major, by course theme or by career pathway, each one answering a different version of “what does this actually mean for me.”
Overview: cutting everything that wasn't a next step
Designed around student actions
I designed the overview so GPA, degree progress, and deadlines all helped students decide what to do next.
Removing features improved clarity
After concept testing, I cut visualizations and secondary panels that distracted from the primary action, making the page more focused and usable.
Enrollment Plugin: linking into iGPS instead of building a sixth tool
Integrated into existing workflows
Connecting Galileo to iGPS turned planning and registration into a single continuous experience.
Visible system feedback
A live progress indicator eliminated uncertainty while schedules were added to the enrollment system.
In Hindsight
The advisor gap remained
Human intervention isn't a layer you unlock after the others, it's a seam that should run through all of them. A student stuck on the simplest task list needs a way to reach a person just as much as a student weighing a fully sourced recommendation.
We chose depth over coverage. Building one core planning experience end to end mattered more than spreading across every part of a student's academic life, including advising, and ending up shallow everywhere instead of strong at one thing.
Reflections
AI suggestions can still come off as prescriptive
Copy alone doesn't guarantee behavior change. If I extended this project, I'd want to test whether the disclosure language actually changed how students weighed a recommendation and explore different interaction patterns for this use case.
Prescriptive
A percentage next to a major reads like an answer, even with a tooltip sitting right next to it explaining how it got there.
Assistive
This pattern already existed a few screens over. Four options with equal weight given to each and tracing back to a specific interest instead of ranked against the others.