Inspiration

Compass was inspired by a simple problem many students face: when everything feels overwhelming, it’s hard to know where to start.

Campuses offer tutoring, counseling, emergency grants, food pantries, advising, health services, and more — but these resources are scattered across different departments and websites. When a student is stressed, behind in class, sick, struggling financially, or feeling isolated, the biggest barrier isn’t the solution — it’s finding the right doorway.

We built Compass to reduce that friction. Instead of navigating institutional menus, students can describe what they’re going through in their own words — and Compass guides them toward the right next step instantly.


What it does

Compass is an intelligent student support navigator that transforms unstructured student concerns into clear, prioritized action steps.

When a user types what they’re experiencing, Compass:

  • Detects relevant support categories (academic, mental health, financial, health, food, housing, etc.)
  • Prioritizes urgency intelligently
  • Generates structured Do now / Next / Also helpful steps
  • Displays campus-specific resources
  • Suggests nearby off-campus options
  • Overrides everything with emergency guidance if crisis language is detected

It can handle layered situations such as:

  • “I’m overwhelmed and failing two classes.”
  • “My rent is due and I can’t cover it.”
  • “I twisted my ankle and can’t walk.”
  • “I feel lonely and don’t know anyone here.”

Instead of overwhelming users with links, Compass provides clarity and direction.


How we built it

Compass uses a rule-based intent detection and ranking system built entirely in JavaScript.

Each message is processed through:

  • Phrase matching
  • Regex pattern detection
  • Weighted intent scoring
  • Urgency boosts
  • Scenario detection (overload, academic recovery, financial instability, physical injury, etc.)

Each detected intent receives a score calculated as:

$$ \text{Final Score} = \text{Match Weight} + \text{Urgency Boost} + \text{Situational Boost} $$

The system then:

  • Ranks the top distinct support categories
  • Limits repetitive scheduling actions
  • Caps recommendation buckets to prevent cognitive overload
  • Blends multiple categories when problems overlap

Campus routing is powered by structured { tag, sub } matching with fallback logic to ensure resources always display correctly.

Nearby support options are generated dynamically using contextual Google Maps queries and optional geolocation.

Everything runs client-side for speed, privacy, and accessibility.


Challenges we ran into

Handling overlapping problems

Students rarely describe a single issue. Someone might be academically behind, financially stressed, and emotionally overwhelmed at the same time. Balancing these without flooding the user required careful ranking and prioritization logic.

Preventing repetitive recommendations

Early versions suggested too many scheduling actions (book tutoring, schedule counseling, call advisor, etc.). We implemented scheduling caps, de-duplication logic, and category diversity rules to keep guidance focused and calm.

Crisis handling

We built a strict override system so that self-harm language immediately surfaces emergency resources and bypasses normal routing.

Debugging runtime issues

We encountered a JavaScript Temporal Dead Zone error caused by referencing a const variable before initialization. Fixing it required restructuring the recommendation pipeline to apply scenario modifiers safely after state creation.


Accomplishments that we're proud of

  • Designing a multi-layer intent detection system without relying on heavy machine learning
  • Creating urgency-aware ranking logic
  • Generating structured, actionable guidance instead of just listing links
  • Ensuring crisis detection safely overrides all other behavior
  • Building a clean, calm user experience that reduces cognitive load

Most importantly, Compass doesn’t just categorize problems — it helps students take their first step.


What we learned

  • Students need prioritization more than information
  • Reducing cognitive load is itself a form of support
  • Thoughtful logic design can feel intelligent without complex ML models
  • Small adjustments in ranking dramatically affect perceived usefulness

What’s next for Compass

  • Machine learning–based intent refinement
  • Integration with campus appointment booking systems
  • Anonymous analytics to identify common student pain points
  • Multi-campus expansion
  • SMS-based quick routing

Our goal is simple:

Reduce the distance between
“I don’t know what to do.”
and
“Here’s your next step.”

Built With

Share this project:

Updates