Inspiration
Finding reliable housing as a student or renter is frustrating and time-consuming. Scattered data, outdated listings, and fake reviews make it difficult to trust what’s online — especially for students moving to new cities or countries.
We wanted to build something that brings transparency, trust, and intelligence into the housing search process. That’s how Room Hunt was born — an agentic AI platform that combines Bright Data web scraping, Gemini LLM reasoning, and Mapbox geospatial intelligence to help users discover verified, well-rated apartments with evidence-backed insights.
What it does
Room Hunt allows users to search for apartments using natural language queries, such as:
“Find apartments under $1,800 within 1 mile of CSU Chico with good management.”
The system automatically:
Understands the query using Gemini LLM.
Scrapes listings from trusted housing sources via Bright Data.
Enriches data with reviews, amenities, and neighborhood information.
Calculates distances using Mapbox and Google Maps APIs.
Analyzes sentiment to evaluate management, maintenance, and safety.
Scores each property (0–100) across five key categories:
Management – reliability & responsiveness
Locality – safety & accessibility
Value for Money – rent vs amenities
Maintenance – repair quality
Commute – proximity to university
Each result is displayed as an Apartment Intelligence Card — showing verified data, category scores, and transparent explanations so users can make confident decisions.
“Find apartments under $1,800 within 1 mile of CSU Chico with good management.”
The system automatically:
Understands the query using Gemini LLM.
Scrapes listings from trusted housing sources via Bright Data.
Enriches data with reviews, amenities, and neighborhood information.
How we built it
Bright Data → multi-source web scraping (Zillow, Apartments.com, Google Maps, Reddit).
Gemini LLM → query interpretation, reasoning, and summarization.
Mapbox + Google Maps API → geocoding, distance, and nearby points-of-interest analysis.
Python (Pandas, NLP) → data cleaning, enrichment, sentiment scoring.
React + Node.js → prototype frontend and API integration.
To quantify apartment performance, we built a weighted scoring model:
Challenges we ran into
Handling dynamic and changing website structures during scraping.
Dealing with incomplete / inconsistent data across platforms.
Managing API rate limits and scraping costs efficiently.
Maintaining LLM explainability — every score required traceable reasoning.
Ensuring smooth multi-API integration without latency bottlenecks.
Accomplishments that we're proud of
Built a fully functional end-to-end AI pipeline — Query → Scrape → Enrich → Score → Summarize.
Seamlessly integrated Bright Data, Gemini, Mapbox, and Google Maps APIs.
Delivered transparent, evidence-backed scoring to promote user trust.
Designed an intuitive UI for apartment filtering, comparison, and visualization.
Demonstrated how agentic AI orchestration can make complex, real-world searches simple and reliable.
What we learned
How to orchestrate LLMs as autonomous agents for reasoning and enrichment.
The importance of cross-source validation in ensuring data reliability.
How geospatial APIs enhance contextual insights for decision-making.
Why explainability and user trust are as vital as technical performance.
What's next for Adrei
AI-based roommate matchmaking — connecting renters with compatible housemates through behavior, lifestyle, and habit analysis.
Personalized scoring that adapts to user preferences (e.g., safety vs price).
Fraud detection and lease document analysis for safer renting.
User profiles and saved searches for ongoing discovery.
Expansion to multiple universities and cities with continuous data freshness tracking.
Our long-term vision is to make Room Hunt the most transparent, personalized, and trustworthy AI housing assistant — helping people find homes and roommates they’ll truly love.
Log in or sign up for Devpost to join the conversation.