Dwllr.ai

A smart, AI‑powered ‘Tinder for housemates’ that matches compatible tenants, connects them with suitable properties, and streamlines the entire renting experience for both tenants and landlords.

Comment

Dwllr.ai: Project Overview

Inspiration

The rental market is fundamentally broken. Current platforms treat properties as isolated units and tenants as mere financial profiles, completely ignoring the human element of shared living. Tenants are often forced to move in with strangers without knowing if their lifestyles, cleanliness standards, or financial habits align. On the flip side, landlords face unpredictable applicant quality and mismatched groups. We wanted to shift the focus from property-first to people-first, creating a system where compatibility is the foundation of the lease.

What it does

Dwllr.ai is an AI-powered rental ecosystem that functions as a matchmaking platform for the housing market.

  • For Tenants: It acts as a "Tinder for housemates." Through a structured onboarding flow capturing personality traits and lifestyle needs, our AI engine forms harmonious groups.
  • For Landlords: It provides a curated stream of pre-matched, compatible tenant groups, reducing turnover and management friction.
  • The Experience: Tenants swipe through recommended homes; once a match is formed, the platform generates a group SMS chat to facilitate immediate coordination and application.

How we built it

The system is built as a modern, scalable web application using a modular service architecture:

  • Frontend: Developed with Vite and React to power the tenant app, landlord portal, and admin dashboard.
  • Backend: Supabase manages authentication, user profiles, property listings, and real-time updates.
  • AI Engine: We integrated Gemini to drive our core recommendation logic. Compatibility is determined using high-dimensional vector embeddings to calculate a similarity score between users based on the cosine similarity of their lifestyle vectors:

$$S(u_i, u_j) = \frac{\mathbf{v}_i \cdot \mathbf{v}_j}{|\mathbf{v}_i| |\mathbf{v}_j|}$$

  • Infrastructure: Background workers handle matching cycles and generate high-intent interest reports for landlords.

Challenges we ran into

The primary challenge was translating subjective human behavior into objective data. We had to iterate on our embedding models to ensure that "personality matching" wasn't just keyword matching, but a nuanced understanding of lifestyle compatibility. Additionally, balancing the marketplace supply (listings) with demand (pre-matched groups) required precise filtering logic to prevent "choice paralysis" for tenants.

Accomplishments that we're proud of

We are proud of our Dynamic Matching Engine, which successfully moves beyond basic filters to evaluate group harmony. We successfully implemented a seamless transition from a digital "swipe" to a real-world communication channel (Group SMS), bridging the gap between discovery and action.

What we learned

We learned that in high-stakes environments like housing, AI must be a facilitator, not a black box. Users need to see why they are matched. We also gained deep insights into using LLMs as reasoning engines for multi-dimensional sorting problems rather than just text generation.

What's next for Dwllr.ai

The future of Dwllr.ai focuses on the full tenancy lifecycle:

  • Identity & Security: Integrating Liveliness Verification and biometric ID checks to eliminate bad actors.
  • Operational Tools: Developing Chore Management modules and automated Rent-Split Logic.
  • Financial Integration: In-app payment modules for holding deposits and monthly rent.
  • Scalability: Expanding the vector space to include more granular lifestyle data for even more accurate matching.

Built With

Share this project:

Updates