Homi
I teamed up with a group of amazing developers at CalHacks 12.0, held at UC Berkeley, to explore how AI agents could reshape everyday digital experiences. Over a weekend, we built HOMI, an AI-powered reimagining of Airbnb that simplifies how people list and discover vacation rentals.
project type
AI Product design
Skill
figma | UX design
Year
2025
Role
DESIGNER
“ Guests browse an average of 17–25 listings before even booking. ”
(Source: Airbnb Report 2023)
“ Guests browse an average of 17–25 listings before even booking. ”
“ Over 80% of Airbnb hosts are individuals, not companies, and many list only one property — they lack time or design skills for professional listings. ”
(Source: Airbnb Report 2023)
“ Over 80% of Airbnb hosts are individuals, not companies, and many list only one property — they lack time or design skills for professional listings. ”
Listings and Findings
Shouldn't Take Hours
Hosting today is still manual, tedious, and emotionless. Most platforms rely on forms, filters, and guesswork
AI that
Understands Spaces
We designed Homi, an AI-powered app that reimagines how people host and discover stays.
For hosts, it automatically generates listings from a simple room scan.
For guests, it offers a way to explore immersive, vibe-based stays through swiping and AR previews.
HOMI
HOMI
An AI-powered home discovery and hosting platform that understands spaces
An AI-powered home discovery and hosting platform that understands spaces
Host
Scan space
Generate listing
chat & refine
Computer vision detects rooms, furniture, and amenities.
AI writes title, description, and price automatically.
Co-create and confirm details through conversation — no forms or spreadsheets.
guest
swipe to discover
Inspiration Scan
AR Preview
Explore homes through visuals and vibe by swiping.
Upload a photo or aesthetic to find matching listings.
Step into spaces virtually and sense the atmosphere before booking.
#DFDED3
#000000
#DFDED3
#000000
Lato
Semi Bold & Regular
Aa Bb Cc Dd Ee Ff Gg Hh Ii
Jj Kk Ll Mm Nn Oo Pp Qq Rr
Ss Tt Uu Vv Ww Xx Yy Zz
#FFF1AA
#AEECFF
#FFC38C
#EAFF6B
#E1A0FF
#FEFBED
#BBB4B6
#1C1D22
Montserrat
Semi Bold & Regular
Aa Bb Cc Dd Ee Ff Gg Hh Ii
Jj Kk Ll Mm Nn Oo Pp Qq Rr
Ss Tt Uu Vv Ww Xx Yy Zz
-$124.18
Total Expense
Relationship of
mélodrame
Regular
Aa Bb Cc Dd Ee Ff Gg Hh Ii
Jj Kk Ll Mm Nn Oo Pp Qq Rr
Ss Tt Uu Vv Ww Xx Yy Zz
#FFF1AA
#AEECFF
#FFC38C
#EAFF6B
#E1A0FF
#FEFBED
#BBB4B6
#1C1D22
Montserrat
Semi Bold & Regular
Aa Bb Cc Dd Ee Ff Gg Hh Ii
Jj Kk Ll Mm Nn Oo Pp Qq Rr
Ss Tt Uu Vv Ww Xx Yy Zz
-$124.18
Total Expense
Guest - Swipe Interaction
Inspired by Tinder-style interactions, we designed a swipeable property card system that lets users quickly express preferences while browsing intelligently matched listings
Guest - AI Agent Designed
to Help You Discover
Discovery starts as a conversation. Guests can either type what they’re looking for or talk directly to the chatbox, describing their needs in natural language.
HOMI’s AI translates these inputs into dynamic filters and curated results, then continues the dialogue by asking thoughtful follow-up questions—clarifying preferences, tradeoffs, and priorities. As users explore, an integrated Q&A assistant provides contextual answers about each space, helping guests move from vague intent to confident decisions with ease.
Host - AR Scanning and Listing
When hosts create listings, our vision system scans uploaded photos to automatically detect amenities. For example, upload a photo of your backyard and it spots the pool, hot tub, and BBQ grill.
Host - Listings and Sellings
By combining vision analysis with market data, HOMI’s AI generates amenity lists and pricing recommendations that reflect both the space and its context. Hosts review, refine, and publish with confidence, supported by an assistant that explains every suggestion
Reflection
Working on HOMI was a rare opportunity to bridge AI with real human decision-making.
As a designer, the most important lesson wasn’t just about building with AI, but about designing with AI as a collaborator, not a replacement. The system worked best when it didn’t assume certainty — when it asked a question, paused for feedback, and adapted. Users didn’t just want a price or a list; they wanted reasoning and control. That insight reframed our product decisions
Listings and Findings
Shouldn't Take Hours
Hosting today is still manual, tedious, and emotionless. Most platforms rely on forms, filters, and guesswork
Guest - Swipe Interaction
Inspired by Tinder-style interactions, we designed a swipeable property card system that lets users quickly express preferences while browsing intelligently matched listings
Guest - AI Agent Designed
to Help You Discover
Discovery starts as a conversation. Guests can either type what they’re looking for or talk directly to the chatbox, describing their needs in natural language.
HOMI’s AI translates these inputs into dynamic filters and curated results, then continues the dialogue by asking thoughtful follow-up questions—clarifying preferences, tradeoffs, and priorities. As users explore, an integrated Q&A assistant provides contextual answers about each space, helping guests move from vague intent to confident decisions with ease.
Host - Listings and Sellings
By combining vision analysis with market data, HOMI’s AI generates amenity lists and pricing recommendations that reflect both the space and its context. Hosts review, refine, and publish with confidence, supported by an assistant that explains every suggestion
Host - AR Scanning and Listing
By combining vision analysis with market data, HOMI’s AI generates amenity lists and pricing recommendations that reflect both the space and its context. Hosts review, refine, and publish with confidence, supported by an assistant that explains every suggestion
Reflection
Working on HOMI was a rare opportunity to bridge AI with real human decision-making.
As a designer, the most important lesson wasn’t just about building with AI, but about designing with AI as a collaborator, not a replacement. The system worked best when it didn’t assume certainty — when it asked a question, paused for feedback, and adapted. Users didn’t just want a price or a list; they wanted reasoning and control. That insight reframed our product decisions
“ Guests browse an average of 17–25 listings before even booking. ”
“ Over 80% of Airbnb hosts are individuals, not companies, and many list only one property — they lack time or design skills for professional listings. ”