Inspiration

Moving to a new city or country as an international student is overwhelming. Traditional relocation tools give you generic, sterile recommendations: “Visit the nearest Starbucks” or “Check out the campus library.” They ignore who you actually are. When a student leaves home, they lose their comforting rituals—the indie record shop where they unwound after exams, the quiet reading café with natural light, or the niche comfort food that grounded them. We asked ourselves: What if an AI could translate your personal cultural fingerprint into your new surroundings before you even unpack your bags? That vision inspired HomeBase—an autonomous Cultural Relocation Agent that transforms cultural displacement into immediate belonging.

What It Does

HomeBase acts as an intelligent landing copilot for relocating students through two core experiences: Settle (The First-30-Days Actionable Landing Blueprint): Synthesizes your taste profile (favorite artists, authors, cinematic styles, comfort dining) and cross-references it with live geospatial nodes around your new campus (e.g., Cambridge / Harvard / MIT). Generates a hyper-personalized, 30-day relocation itinerary divided into cultural phases: Arrival & Cultural Anchor, Acoustic Focus & Exploration, Niche Subcultures, and Community Integration. Each recommended spot features an interactive Leaflet/OpenStreetMap view with verified walking distance, direct Google Maps navigation links, and an explicit Taste Link explaining the cultural bridge (e.g., "Matched through your affinity for Radiohead and ambient post-rock workspaces"). Connect (Taste-Affinity Peer Matching & Real-Time Chat): Eliminates awkward cold introductions by pairing newly arrived students based on cultural resonance, shared comfort tastes, and academic focus. Shows live online presence indicators and an "All vs. Online Now" toggle. Suggests mutual meetup spots within walking distance from both peers. Includes a built-in 1-on-1 direct messaging space equipped with cultural icebreaker prompts to break the ice instantly.

How We Built It

Taste Intelligence Engine: Powered by the Qloo API (/v2/insights, /v2/entities), querying multidimensional taste graphs across music, literature, dining, and cultural affinities to compute taste vectors. Fast Reasoning Agent: Integrated Groq running llama-3.3-70b-versatile with low latency to synthesize real-time itineraries, explain affinity bridges, and reason over student constraints. Geospatial & Mapping Layer: Leveraged OpenStreetMap & Nominatim for live coordinates and geospatial geocoding, rendered with Leaflet for interactive campus mapping. Taste Onboarding: Seamlessly ingests student preferences manually or through the Spotify Web API for automated music taste extraction. Backend & Persistence: Built on Next.js 15 App Router and TypeScript, backed by MySQL (Aiven Cloud) with automatic schema initialization and an offline JSON store fallback. Deployment: Hosted on Vercel with a serverless edge architecture.

Challenges We Ran Into

Cross-Domain Taste Translation: Bridging disparate cultural domains—like mapping an appreciation for Haruki Murakami novels to a specific aesthetic café in Cambridge—required careful prompt engineering and structured reasoning constraints on top of Qloo’s entity vectors. Map Rendering in Server-Side Frameworks: Leaflet relies on the browser's window object. Integrating interactive dynamic maps in Next.js SSR required custom dynamic imports and SSR hydration workarounds. Zero-Friction Reliability: Ensuring the app gracefully handles database connectivity hiccups during demo judging led us to engineer an automatic self-healing database schema creator alongside a resilient persistent fallback store.

Accomplishments That We're Proud Of

Deterministic Taste Bridges: We didn't build just another chatbot. HomeBase explains why a venue was selected with explicit cultural links that feel genuine and deeply personal. Zero Seed Data in Connect: We replaced static demo mockups with a live peer-matching system that connects real students registered on the platform in real time. Speed & Flow: Groq’s high-speed inference combined with Qloo’s taste graph creates a sub-second itinerary generation experience.

What We Learned

Taste is fundamentally relational: finding familiar comfort in an unfamiliar city is less about finding exact replicas and more about finding places with matching atmospheric and aesthetic sensibilities. How to architect robust LLM agent loops that consume structured third-party graph API data (Qloo) and ground their responses strictly in verified geographic coordinates.

What's Next for HomeBase

Group Taste Mesh: Organizing taste-matched micro-cohorts (e.g., a 4-person weekend vinyl-listening or study club) instead of only 1-on-1 pairings. Campus Housing Integration: Helping students select university dorms or apartments based on cultural affinity heatmaps of neighboring districts. Local Business Partnerships: Enabling student-friendly indie bookstores and independent roasteries to offer targeted student discounts to culturally matched newcomers.