top of page
ASSISTIVE DEVICE · LLM · PRODUCT DESIGN
Enabling the low vision community to relive memories through touch, sound, and smell
.png)
ROLE
Product Designer
TIMELINE
February - April 2026
TEAM
3 Designers
1 Engineer
SKILLS
Product Design
User Research
Vibe Coding
Hardware Prototyping
anchorOverview
OVERVIEW
Most assistive technologies for the blind are need-based and often overlook social and emotional inclusion.
As part of a multidisciplinary team, I designed EMORA, a Multi Sensory Memory Interface using NLP that enables the low vision community to relive memories through touch, sound, and smell.
EXISTING TECH FOR THE BLIND
Conventionally, memories are shared through...
But what if we could share memories through...?
So we asked:
How might memory-sharing evolve if accessibility shifts from screen-based consumption to shared tangible experiences?
WHAT IS UNCERTAIN?
THE BRIEF
anchorBrief
How might we design a multi sensory memory ecosystem that empowers the visually impaired to capture, preserve and share life experiences through touch, sound and smell, transforming fleeting moments into tangible artifacts?
anchorSolution
SOLUTION
Emora: the device that stores memories through scent and audio...
...using NLP for the visually impaired
Person speaks a memory from which keywords are extracted. A texture tile is fabricated according to the memory. When the same tile is inserted into the device, texture + sound + scent triggers the memory.
RESEARCH
anchorResearch
Tactile perception works best with clear spatial and physical differences
Blind and visually impaired users distinguish textures most effectively when they have clear spatial structure, strong physical contrast, and consistent patterns, while fine details and subtle textures should be avoided to reduce confusion.
raised patterns
ridges
bumps
rough vs smooth
structured grids
strong material contrast
fine details
subtle textures
visual drawings
irregular patterns
anchorProcess
PROCESS
Utilizing NLP to Translate Emotional States into Physical Tactile Parameters
The Natural Language Processing system is a translator between human language and a curated keyword system. The library was rebuilt from scratch around three dimensions that map directly to what a 3D printed surface can reproduce and what a blind person would naturally reach for in language.
Stage 1
Providing context to the agent and setting up a keyword library for it to refer to
Stage 2
Setting up rules to provide structure to the agent's responses.
Stage 3
Asking the agent to match tile textures to the right keywords from the library.
ROLE OF THE AGENTIC SYSTEM
_edit.jpg)
anchorAgent
AGENTIC SYSTEM
How are memories stored and retrieved?
The first step involved extracting keywords and matching textures to the memory from the audio input. The library was rebuilt from scratch around three dimensions that map directly to what a 3D printed surface can reproduce and what a blind person would naturally reach for in language.
The next step involved using an image texture analyzer to retrieve memories through audio and scent using the 3D printed texture tiles. Each texture is mapped to a particular audio and scent which is activated using computer vision to activate the memory.
anchorPrototype
HARDWARE PROTOTYPING
What would this device look like?
The next phase was to actually build a device which could retrieve these memories and be accessible to all users while being screen-free. Many forms were 3D printed and tested to understand what would work best.
PROCESS
Different textures and sizes were analyzed to see which would work best. How course or smooth should the textures be? Can the textures behave a buttons that get activated when pushed down?
Introducing scent modules which are activated when the texture tiles are pushed down, behaving as buttons.
TESTING
How does it work?
Iterating with Feedback!
We got users from the low vision community to test our product and received feedback on what worked, what was surprising and what could be improved.
KEY INSIGHTS
“I wonder if I could choose my own textures, sounds, or even scents, something connected to memories that are personal to me?”
- Chris, Lighthouse
“When I remember a moment, I don’t just remember one object. I remember the environment around it. The sounds, the smell, the people, the feeling of being there.”
- Corvin, Lighthouse
anchorTakeaways
TAKEAWAYS
Challenges I faced...
Distinguishable Texture Selection
Choosing textures that are easily distinguishable through touch while still being meaningful enough to represent different memories.
LLM as Component vs Design Tool
Defining that the LLM should function as a prototype component (processing memory input) rather than as a design aid (shaping the system).
Keyword Variable Selection
Identifying the right variables (surface quality, resistance, structure) to accurately map audio memories to meaningful sensory outputs.
Looking Forward...
Emotional Communication for Blind Children with Autism Spectrum Disorder
Multi sensory tangible devices could help children recognize, express, and communicate emotions through textures, scents, and audio cues supporting social interaction and emotional regulation.
Memory Recall Support for Older Adults with Dementia
Tactile memory artifacts combined with familiar scents and sounds could help trigger autobiographical memories, emotional comfort, and meaningful connection with caregivers and family.
bottom of page