NYC RENTAL DISCOVERY PLATFORM

Explorentory

AI  ·  Machine Learning  ·  Geospatial Visualization

01

How can an AI-assisted experience translate natural language preferences into personalized rental recommendations?

02

How can machine learning adapt rental recommendations based on user behavior and implicit feedback?

03

How can interactive visualizations clarify housing information and make spatial trade-offs legible?

Zillow

Zillow · Traditional Platforms

Explorentory

Explorentory

  • Hard filters — all-or-nothing, binary inclusion
  • Miss properties slightly above your budget
  • No visualization of trade-offs across options
  • Manual one-by-one property review
  • Explores ± range beyond your initial filters
  • Learns your specific, ambiguous preferences
  • Visualizes trade-offs spatially and across axes
  • AI-assisted exploration and comparison

PostgreSQL
+ PostGIS

~4M synthetic NYC units
260 neighborhood polygons
GeoJSON boundary data
Geospatial indexing

FastAPI
REST →

Backend

6 REST endpoints
OLS regression (scikit-learn)
OpenAI Responses API
PostGIS spatial queries

JSON +
GeoJSON →

Vanilla JS
Frontend

MapLibre GL polygons
Canvas-based charts
Client-side LLM filtering
Snapshot state history

01 Initial Input
→
02 Neighborhood Selection
→
03 Property Rating Survey
→
04 Explore & Discover
Step 1 — Preferences
01 Preferences & Priorities
Step 2 — Neighborhood
02 Neighborhood Selection
Step 3 — Rating
03 Property Rating Survey
Step 4 — Results
04 Results Panel & Chat

~4M

Properties
in Database

Rule-Based
Filter

→

Rent ±20%/+5%
BD ±1 · BA ±1

~100K

Candidates
Shortlisted

Rule + ML
Scoring

→

50% rule weight
50% OLS regression

5,000

Top Results
Surfaced

AI-Assisted
Exploration

→

Chat · Map
Cards · Radar

You

Explore,
Compare & Decide

Three Methods of Machine Decision-Making

1960s – 1990s

Rule-Based
Filter

  • Deterministic & binary
  • Transparent and fast
  • Cannot handle subjectivity

2000s – 2010s

Classical ML
(OLS Regression)

  • Learns patterns from data
  • Requires structured input
  • Handles some ambiguity

2020s →

AI Assistive
Exploration

  • Natural language interface
  • Learns subjective preferences
  • Handles deep ambiguity