Role:
Lead Product Designer
Role:
Lead Product Designer
Role:
Lead Product Designer
Role:
Lead Product Designer
Role:
Lead Product Designer
Date:
2019
Date:
2019
Date:
2019
Date:
2019
Date:
2019
Tags:
Machine Learning, Real Estate
Tags:
Machine Learning, Real Estate
Tags:
Machine Learning, Real Estate
Tags:
Machine Learning, Real Estate
Tags:
Machine Learning, Real Estate
Summary
Summary
Summary
belowmrkt applies machine learning to property search, adapting filters to each user from the very first interaction. It learns fast, feels personal and makes discovery effortless.
belowmrkt applies machine learning to property search, adapting filters to each user from the very first interaction. It learns fast, feels personal and makes discovery effortless.
belowmrkt applies machine learning to property search, adapting filters to each user from the very first interaction. It learns fast, feels personal and makes discovery effortless.
the challenge
the
challenge
Teaching AI to learn from little
belowmrkt set out to make property search smarter through adaptive filters that learned from behaviour. The challenge wasn’t scale but scarcity. Building a machine learning system that could understand user intent from just a few interactions. It needed to feel responsive and personal without relying on big data or long histories.
the challenge
Teaching AI to learn from little
belowmrkt set out to make property search smarter through adaptive filters that learned from behaviour. The challenge wasn’t scale but scarcity. Building a machine learning system that could understand user intent from just a few interactions. It needed to feel responsive and personal without relying on big data or long histories.
the constraints
the
constraints
Small data, smaller window
The team had weeks, not months. Data volume was limited and the model had to start learning from the first session. There was no room for complex onboarding or training loops. The product needed to show intelligence quickly while keeping the interface simple enough to launch fast.
the constraints
Small data, smaller window
The team had weeks, not months. Data volume was limited and the model had to start learning from the first session. There was no room for complex onboarding or training loops. The product needed to show intelligence quickly while keeping the interface simple enough to launch fast.
the process
the
process
Designing learning into interaction
We focused on how people naturally explore properties. Quick scans, shortlists and revisits. Each of those actions became input for the learning model. I led the design direction and shaped the product behaviour around these micro-patterns, working with engineers to define how filters adapt in real time. The interface stayed minimal by design: every gesture carried meaning and every scroll improved the next result.
the process
Designing learning into interaction
We focused on how people naturally explore properties. Quick scans, shortlists and revisits. Each of those actions became input for the learning model. I led the design direction and shaped the product behaviour around these micro-patterns, working with engineers to define how filters adapt in real time. The interface stayed minimal by design: every gesture carried meaning and every scroll improved the next result.
the principles
the
principles
Machine learning made human
The intelligence stayed invisible. Instead of sliders or manual weighting, filters refined themselves as users browsed. The design avoided technical language and emphasised feedback. Subtle animations and layout shifts that showed the system was listening. The goal was to make the app feel intuitive, not algorithmic.
the principles
Machine learning made human
The intelligence stayed invisible. Instead of sliders or manual weighting, filters refined themselves as users browsed. The design avoided technical language and emphasised feedback. Subtle animations and layout shifts that showed the system was listening. The goal was to make the app feel intuitive, not algorithmic.
the reflection
the
reflection
Invisible intelligence
belowmrkt kept its intelligence out of sight. The model adapted quietly while users focused on discovery, not data.
the reflection
Invisible intelligence
belowmrkt kept its intelligence out of sight. The model adapted quietly while users focused on discovery, not data.
Impact
Impact
Impact
belowmrkt launched quickly and gained early traction. Users spent less time adjusting filters and more time exploring listings that matched their intent. The adaptive system improved engagement metrics week by week without extra data collection. A lightweight build proved powerful enough to shift expectations for property discovery.
belowmrkt launched quickly and gained early traction. Users spent less time adjusting filters and more time exploring listings that matched their intent. The adaptive system improved engagement metrics week by week without extra data collection. A lightweight build proved powerful enough to shift expectations for property discovery.
belowmrkt launched quickly and gained early traction. Users spent less time adjusting filters and more time exploring listings that matched their intent. The adaptive system improved engagement metrics week by week without extra data collection. A lightweight build proved powerful enough to shift expectations for property discovery.
About me
About me
I’m Öncel, a multidisciplinary designer who creates products, systems and experiences people rely on every day. I’ve partnered with global brands and startups at different stages, taking ideas from zero to MVP, leading complex designs that reached millions of users and building foundational design systems that grow with teams.
I'm based in APAC
I'm based in APAC
Singapore, Bangkok
Singapore, Bangkok
Say hi
Say hi
Got something on your mind?
Just say hi. I reply quickly.
Got something on your mind?
Just say hi. I reply quickly.
Got something on your mind?
Just say hi. I reply quickly.