Bonded

Inspiration College can be one of the loneliest times in a person’s life. Between new environments, social pressure, and the overwhelming noise of digital life, students struggle to form meaningful relationships beyond surface-level interactions. We built Bonded to help students form authentic friendships and romantic connections by focusing on emotional compatibility, shared values, and lifestyle alignment.

What it Does Bonded is an AI-powered platform that helps students find genuine connection through two complementary experiences:

Friend Mode

Designed for meeting roommates, study partners, and friends.

Students answer short lifestyle questions (“Are you a night owl or early riser?”) and select at least three interests per category such as music, fitness, gaming, or food.

Gemini evaluates lifestyle compatibility, shared interests, and communication styles to recommend potential friends and roommates who are likely to get along.

Prioritizes comfort, relatability, and belonging over appearance.

Love Mode

A slower, more intentional flow that emphasizes emotional connection before physical attraction.

Starts with a personality and emotional quiz that builds each user’s unique Love Print—a profile of empathy, humor, and communication style.

Gemini uses the Love Print and user ratings to balance 70% emotional compatibility and 30% attraction.

Users progress through a guided, trust-building experience:

Level 1: Blind text-only chat.

Level 2: Voice messages unlocked.

Level 3: Photos and profiles revealed.

Level 4: Optional in-person or video meeting.

Throughout the process, Gemini powers the Link AI companion that helps users reflect on their emotions and conversations in real time.

How We Used Gemini Gemini is central to Bonded’s intelligence and emotional integrity.

Smart Matching and Personality Reasoning

Gemini analyzes users’ lifestyle tags, interests, and Love Prints to understand how compatible two people might be.

It ranks matches not just by shared traits, but by how those traits interact, identifying complementary or balanced personalities.

In Friend Mode, Gemini prioritizes lifestyle and mutual interests. In Love Mode, it blends emotional resonance with measured attraction.

Message Screening and Safety

Gemini acts as an AI moderator, reviewing every message in real time.

It flags or blocks messages that contain personal contact information, social media handles, or inappropriate language.

This ensures the platform remains safe, authentic, and free from spam or harassment.

Emotional Guidance with Link AI

Gemini supports the “Link” feature, which provides reflection prompts like “How did that conversation make you feel?”

These prompts help users develop emotional awareness and encourage healthy digital communication habits.

Dynamic Recommendations and Adaptation

Gemini continuously learns from user feedback and engagement patterns, refining match suggestions over time.

This makes Bonded’s recommendations increasingly personal and accurate.

How We Built It Frontend: React Native (Expo) for cross-platform deployment.

Backend: Supabase for authentication, database management, and real-time chat.

AI Integration: Gemini API for reasoning, moderation, and emotional insights.

UI: TailwindRN and Framer Motion for smooth, emotional, human-centered animations.

Carousel: Skiper48 swipe deck for intuitive friend and match discovery.

Matching Engine: Hybrid weighted algorithm combining interests, lifestyle, and emotional metrics.

Challenges We Ran Into Getting Gemini to handle both moderation and ranking logic seamlessly.

Designing a quiz flow that felt personal and meaningful without overwhelming users.

Balancing Friend Mode’s instant gratification with Love Mode’s intentional pacing.

Maintaining natural, fluid conversations even with strict moderation safeguards.

Accomplishments We’re Proud Of Created a dual-mode system that treats both friendship and dating as valuable human experiences.

Developed a working Gemini-powered moderation and reasoning pipeline.

Built a dynamic onboarding-to-match pipeline that personalizes every journey.

Delivered a complete and emotionally intelligent app in under 72 hours.

What We Learned Emotional design is as important as technical performance when building social apps.

AI moderation can strengthen user trust when implemented transparently.

Compatibility extends beyond shared interests—it’s about emotional rhythm and lifestyle balance.

What’s Next for Bonded Launch pilot programs across campuses starting with the University of Rhode Island and Boston-area schools.

Expand Link AI to include personalized post-date emotional feedback and coaching.

Add verified roommate-matching for on-campus housing systems.

Integrate event-based discovery for clubs, sports, and local student communities.

Partner with universities to provide mental health and connection-building resources directly through the app.

Creativity & Originality Bonded introduces a fresh take on digital connection by combining friendship, dating, and emotional reflection in a single AI-driven experience. The progressive reveal and dual-mode system are entirely new to student networking.

Use of Gemini Gemini is leveraged in three ways—smart matchmaking, conversation moderation, and emotional guidance—showcasing reasoning, prompt design, and automation capabilities. It’s used as both the brain and conscience of the app.

Impact & Relevance Addresses the loneliness epidemic among students by providing a safe, trusted, and emotionally supportive way to connect. Bonded creates belonging on campuses where social connection is often missing.

Learning & Execution The team built an end-to-end app using modern tools and successfully integrated Gemini for reasoning and safety, demonstrating curiosity, adaptability, and rapid technical iteration.

Presentation & Clarity The concept, purpose, and technical process are communicated clearly through the demo and write-up, illustrating both emotional intent and AI sophistication in a way that is easy to understand.

Built With

• gemini
• react-native
• supabase