Roomies

The landing page of Roomies.
Authentication page to log in.
Page to do matches.
Chat with matched people.
Add preferences so that the AI can provide more tailored suggestions.
Edit your profile to better match you.

Inspiration

With the housing, living, rental, everything crisis imploding in our day-to-day life, we've seen people struggling to make ends meet and scraping the bottom trying to survive.

In particular, being students surrounded by people who live out of home, we've seen the difficulty of finding a compatible roommate/share house mates especially as most "good", affordable places are usually 3+ bedroom houses -- afterall, a good sharehouse mate or roommate, makes or breaks a rental experience. Common techniques to tackle this problem includes spamming instagram stories looking for roommates/places, going through an arduous (or even minimal) number of friends to see if they want to rent with them, or even the borderline useless Flatmates platform.

We wanted to create the bridge between renters so that they can 'vibe check' their potential roomies before actually signing the lease with them.

What it does

A website akin to Hinge where instead of finding dates, you swipe for whether you would like to possibly share an accommodation, except every suggestion is tailored for you! After sending the 'like' and the other person approves, you are able to chat with them directly to discuss common interests, needs, and perhaps even organise a hang out. This is meant to encourage a space to detect if they are a potentially 'good' roomie.

How we built it

Frontend: Typescript, React, HTML/CSS, Bootstrap, and Tailwind Backend: Typescript, Langflow API, REST API, Express.JS, Node.JS, MongoDB, Astra DB

User provides their preferences and background/important information, and this is vectorized and stored in the AstraDB. The matching system which creates tailored suggestions to the user is built using Langflow on Datastax, and a fetch call is made to identify the most tailored suggestions.

Challenges we ran into

Overall: As this is an ambitious project with many components over restricted deadline, we were not able to implement authentication capabilities.

Frontend: Had difficulty of maintaining compatible styling, as each individual working on frontend had different preferences.

Backend: Initially wanted to use Javascript due to compatibilities with the current stack, but because we wanted to use a vectorised database on AstraDB for better performance of the model, we had to use Typescript. We then decided to use Typescript rather than Javascript.

Accomplishments that we're proud of

Fully navigational navigation bar.
Design is peaceful and delivers idea concisely, because we know that finding people to live with is stressful.
Has a working backend where we can successfully fetch matching existing users to current user from Langflow API.
Landing page works (we were stuck on how to place our Spline design which always had an awkward position on the page whenever changing screen size).
Utilising new frameworks that we were not comfortable with.

What we learned

Frontend is held by a bluetack and a paperclip.
Learnt what a vectorized dataset is (never heard of it before...).
If you use body or h1, h2, etc in your CSS file it affects every component inside your page and their styling (a miserable lesson learnt...).
How nice, simple, and functional React components are.
A lot of debugging goes towards frontend.
The importance of managing git branches properly.