Beginners in sport progress unevenly, get no feedback in the moment and rarely have access to a coach. Most wearables answer that with more data and more features, which overwhelms them. We asked how a wearable that fits into existing gear could give beginners intuitive feedback and personal coaching.
What I did
Ran the brainstorming sessions (Crazy 8s, Lotus Blossom) that produced the first concepts.
Wrote the interview protocol for semi-structured interviews with 5 people who play sport.
Turned the interview findings into the user persona the team designed for.
Led the ideation and design of the companion app, from user flow and low-fi wireframes to the final screens.
Tested the low-fi app with an interview participant and with stakeholders.
What the team did
Two teammates built the Arduino sweatband prototype with motion sensors and a vibration motor.
We chose the final concept, the feedback pattern and the visual guideline together.
Key decisions
A sweatband, not a helmet or smart clothing.I dropped my two earlier concepts because people would not wear them every day and coaching did not fit into them naturally. Players already wear sweatbands, so there is nothing new to get used to.
Correct on the front of the wrist, incorrect on the back.Interviewees wanted feedback on mistakes to be instant and unmistakable. Two separate positions let a player tell right from wrong without looking at a screen.
A haptics tutorial inside the app.Not everyone knows how to read vibration feedback. In the low-fi test, the participant said the 3D hand made it clear what to do, so the tutorial stayed and became part of onboarding.
Cables and Wizard of Oz when Bluetooth failed.The sensor and Bluetooth connection were unreliable. Instead of losing the test round, the team wired the sensors directly and simulated the feedback by hand.
The result
A tested concept in two parts: a sweatband prototype that gives vibration feedback during play, and an app prototype with personal coaching tips, session analytics, progress tracking and the haptics tutorial.
How the feedback works.Low-fi screens with test feedback.Final app: feedback and analytics.
Limits and next step
The haptic patterns were tested with a simulated prototype, not on court with working hardware. The next step is testing them with beginner players during real practice.
Skills shown
User interviewsPersonaCo-creation workshopsConcept evaluationUser flowsWireframingPrototyping in FigmaUsability testingWizard of OzWorking with hardware teammates
Kumo Compass
A wearable-first travel companion that gives visitors to Japan gentle etiquette guidance at the place and moment it matters.
My roleUX Researcher, UX/UI & Brand Designer (solo)
TypePersonal project
Duration3-week sprint
PlatformsiPhone and Apple Watch
ToolsFigma, FigJam, coded prototype (React)
Kumo Compass (雲コンパス): one short, friendly etiquette tip at the right place and moment.
See it working
A 40-second recording of the coded prototype. After the Figma design, I turned Kumo Compass into a working web app (React and TypeScript, built with Claude as my coding assistant) to test whether the idea holds up outside a click-through mockup.
It runs in the browser, can be added to a phone's home screen, and uses Kyoto as sample content: eight places, each with its own story and manners. A built-in walk simulator lets anyone try a Kyoto walk without being in Japan.
Start a walk. One tap on the home screen. Kumo follows your position, and your location never leaves the phone.
Getting close. As you near a temple, shrine or station, you get one short reminder of the manners for that kind of place.
Arriving. A brush circle marks "destination reached", with the place's story and the two or three manners that matter most there.
It stays calm. At most one tip per place and no more than six an hour, with quiet hours. You choose the pace: Calm, Normal or More.
Map and memories. A woodblock-style map of Japan works offline. Places you reach get a stamp, and each walk is saved on the phone.
The problem
Many travellers want to take part in another culture respectfully and are quietly afraid of getting it wrong. Travel apps help them navigate, but they do not explain local etiquette or why it matters, and never at the moment the traveller needs it.
What I learned
Insight
The barrier is fear, not lack of interest. In the studies I read, most tourists worry about "doing the wrong thing" in an unfamiliar culture.
So I designed
Every tip is written as reassurance: what to do and why it is appreciated, never a warning.
Insight
Worry makes people skip experiences. Some travellers avoid a temple or a local restaurant instead of risking a mistake.
So I designed
Guidance arrives just before the place, and the arrival moment starts with a welcome, so going in feels safe.
Insight
A phone in the hand gets in the way. My persona navigates by phone but does not want to block a crowded street or film where it is not welcome.
So I designed
One short note that can be read at a glance, on the wrist first, and no more than a few an hour.
What I did
Built the research base myself: desk research, mind maps, user flows and the persona Alex Tanaka.
Designed "Enjoy the Moment", a notification system triggered by location that gives short, friendly guidance near cultural sites.
Prototyped a zoomable, interactive map in Figma.
Built a coded prototype of the phone app (React web app) with real location triggers: an outer zone for a gentle reminder and an inner zone for the arrival moment.
Designed the Apple Watch screens following Apple's watchOS Human Interface Guidelines, my first time designing for the wrist.
Created the name, logo and visual identity, and applied them across phone and watch.
Asked for feedback in UX communities during the sprint and worked it in.
Key decisions
Reassure, don't warn.The persona's main feeling was anxiety about offending people. So each notification is written as helpful context, such as a reminder to stay quiet on public transport on the way to Kinkaku-ji, not as a rule to obey.
The watch is a main screen, not an add-on.I used the system notification and map layouts, respected the safe areas and sized text to be read at a glance while walking. The traveller never has to take out a phone.
Working around a creative block.The idle screen stalled after several rounds of wave animations. I switched to designing the "Where You Are" screen, and that gave the idle screen its direction.
Adding branding mid-sprint.Branding was not in the plan. I added it when the screens needed one voice. The logo moved from a sun and compass to a compass softened by a cloud, and the name is easy to say for Japanese and English speakers.
Writing the tips
The words are the product, so I wrote them to a small set of rules. The prototype holds 33 tips for Kyoto and for Japan in general.
伏見稲荷大社 · Fushimi Inari TaishaGates with a story
Every torii here was donated, many by local businesses. Admire them without leaning on or climbing them.
Title: 4 words · Body: 2 sentences, 103 characters
Why first, then what. The reason comes before the request, so it reads as context.
Short enough for a glance. Titles of six words or fewer; bodies of two sentences and under 150 characters.
Say what to do. "Keep voices soft" instead of "Don't be loud".
One idea per tip, and never the same tip twice on one walk.
Checked against sources. My research notes, verified with Kyoto's official tourism site and established Japan travel guides.
The result
Two prototypes and one identity. An interactive Figma prototype across phone and watch: location-based notifications, a zoomable map of Kyoto and an idle screen. And a coded prototype of the phone app that reacts to real location, with a brand that holds on both devices.
Branding
The identity pairs a compass (direction) with a cloud (kumo: adapting gently). Colours carry traditional Japanese names, the typefaces pair a calm monospace with a Mincho for Japanese titles, and the patterns are classic wagara.
Brand sheet: logo, colour, type, pattern and crest.
The "Stay quiet" notification.Nearby landmarks on the watch.Screen iterations with my notes.Branding exploration.
Accessibility
Reduced motion. With the system setting on, drifting petals, the ink transition and the logo movement stop.
Text size. All text is set in relative units, so it follows the reader's browser or phone text-size setting.
Contrast of main text. Dark text on paper measures 12.4:1 and light text on the ink cards 10.7:1, both above the 4.5:1 guideline.
Not by colour alone. Place types use an icon and a label, and visited places get a stamp, not just a colour change.
Limits and next step
This was a self-directed design sprint, so the prototype has not been tested with travellers yet. The next step is a usability test of the notifications with people planning a trip to Japan.
Skills shown
Desk researchPersonaUser flowswatchOS designApple HIGInteractive prototyping in FigmaUI designBranding and logo designSelf-directed learning
Deceptive by Design
A card game and website that teach designers and students to spot dark patterns and see what they cost the planet.
My roleConcept & Card Game Designer, UX/UI Designer, Workshop Facilitator
Team2 designers
Duration2 months
ContextMaster Digital Design, Hogeschool van Amsterdam
Research29-person survey, 5 rounds of playtests and critiques
The printed cards in play at the Feedback Fest.
The problem
The brief asked how digital design choices affect our carbon footprint. Dark patterns such as autoplay and infinite scroll keep people on screens longer, which costs energy, and most people do not notice them. In our survey, only 55% of 29 participants knew the term "digital sustainability" and 7% named environmental impact as a main concern. Neither designers nor users had an easy way to learn to recognise these patterns.
What I did
Came up with the core "Deceptive by Design" concept: framing dark patterns as the link between interface design and sustainability.
Designed the cards through four stages with my teammate, from a crowded first version to a clear trading-card layout.
Designed the website's landing page and content blocks, cutting clutter and raising contrast.
Organised and ran the card workshops and playtests.
Kept the team on track: weekly task tracking, shared documentation and summarising decisions so everyone agreed.
Ran a 5-day self-experiment on my own exposure to dark patterns to test our assumptions.
What we did together
Diego Gunther and I ran the survey, wrote 19 pattern cards and 20 scenarios, and built the website.
Key decisions
Testing before the idea was ready.We ran a public workshop with a rough fill-in-the-blank game. Players laughed, then recognised their own habits. That reaction told us to write scenarios people know from daily life.
Less text, familiar structure.Playtesters liked the tone but got lost in the text. We cut copy, moved to a title, image, effect layout like trading cards, showed points clearly and added a sample card with a practice round.
Simplifying the rules after round five.Testers found the scenario plus pattern combination confusing. We simplified the game, added a point system and colour-coded each card to its pattern category on the website.
Keeping the humour.Some feedback asked for a more precise, serious tone. We kept it light because the goal is discussion, and the playtests showed the jokes are what start it.
The result
A printed card game and a companion website that players reach through a QR code on the cards. Each card names a dark pattern, defines it and states its environmental impact. Our lecturers want to use the game with next year's Master Digital Design students.
Final cards, colour-coded by category.Card colours match the website's categories.Website: stage 3 to final design.
Limits and next step
Some website sections are unfinished and text-heavy. Before the game is used with a new group of students, I would break that text into shorter pieces and finish those sections.
Skills shown
Survey researchConcept developmentWorkshop facilitationPlaytestingIterating on feedbackGame and card designWeb UI designUX writingProject coordinationTeamwork under pressure
AI & Emotional Intelligence
Can AI help people stay respectful when online discussions get heated?
My roleUX Researcher & Designer
ClientCivic Technology
Duration5 months
ContextBA graduation project
FocusResearch, UX, AI
Final design: the AI extension on desktop.
The problem
Publieke Dialogen is an online platform where citizens and municipalities discuss local issues. It was still being built. Online discussions often turn negative, and people misread each other's tone.
Civic Technology asked me to find out whether AI could help keep these discussions respectful, and what could go wrong.
What I did
Read existing studies on how emotions spread in online discussions.
Interviewed 7 people about arguing online and trusting AI.
Let an AI host a real group discussion, to see where it helps and where it fails.
Ran a workshop where participants came up with ideas together with me.
Designed two AI features and tested them with users.
Key decisions
Design for tone, not only for speed.The client wanted faster discussions. My research showed that discussions fail when the tone goes wrong. So I designed for how people say things.
Test the AI before designing anything.The AI was good at starting conversations and giving facts. It was bad at keeping answers short and at handling silence. I designed around those limits.
AI summaries that people can switch off.Summaries helped testers catch up quickly. Some testers found them too general, so I made them optional.
"Mood of the Chat".It shows at a glance whether a discussion is calm or heated. Every tester found it useful and rated it between 4 and 4.5 out of 5.
The result
A design for an AI helper inside the platform, with two tested features. Civic Technology also received my research on what AI moderation can and cannot do.
Framing the problem.Aggregated empathy maps from the interviews.
Want more? The research in detail
How I framed the problem
The project did not start with a single problem. Many directions were possible, so I chose to explore the core principles of human communication: how people interact and what makes an exchange positive or negative.
The spark was the concept of human communicative competence, the ability to convey messages effectively and appropriately. It made me look deeper into how a focus on appropriateness leads to smoother and more productive conversations.
Secondary research: emotion in online discussions
Findings from existing studies of online comment threads.
Prevalence of emotions. 90.6% of comments expressed emotions, and negative emotions heavily outweighed positive ones (only 9.4% were positive).
Emotions and comment types. Positive emotions are more likely in comments that agree (44% positive) and in neutral statements (11% positive). Negative emotions dominate disagreements, provocations and insults.
People respond with the same emotion. Individuals are more likely to express emotions already present in the conversation.
Positive emotions encourage connection. Positive emotions can foster positive responses and strengthen connections between participants.
Negative emotions depend on context. Anger and sadness lead to different responses, and current sentiment analysis struggles to tell them apart.
Thread length follows emotion. Threads with stronger initial emotions tend to be longer, and threads of any length end with similar negative sentiment, an "emotional convergence".
Interviews: six themes
Privacy concerns. All participants had some concern about privacy online, which influenced their preferred platforms and ways of communicating.
Emotional influence. Each participant acknowledged that others' emotions influence their own feelings and reactions.
Emojis for tone. Many participants use emojis to clarify tone and intentions in text.
Scepticism towards AI. Participants saw potential benefits but doubted AI's ability to manage emotions, foster inclusivity and understand social nuance. They emphasised human oversight and transparency.
Inclusivity and authenticity. All participants wanted inclusive, authentic and respectful discussions, with human mediation and personal autonomy.
Managing negativity. Strategies included flagging offensive language, giving warnings and encouraging constructive dialogue.
Experiment: AI as discussion facilitator
The experiment used an AI to facilitate online discussions about cultural topics. The findings show the benefits of AI in initiating discussions, providing information and promoting cultural understanding. Challenges remain in the length and focus of AI responses, in handling participant silence with nuance, and in managing discussion fatigue.
AI can be a valuable asset in starting conversations and prompting participation. Improving the focus of its responses and its understanding of conversational dynamics is crucial for productive online interactions.
Co-creation: what participants asked for
The challenge: how might we design online platforms that encourage users to reflect on their emotional state before posting a response in a heated conversation?
Feeling heard. People want to express their emotions clearly, using emojis and explanations, and want time to reflect before responding.
Respectful chat. Everyone wants to feel safe and respected: clear rules, like no bullying, and tools to communicate politely, like choosing a friendly tone.
Clear communication. Pictures and summaries help people understand each other. Confirmation before posting and easy-to-read notifications prevent misunderstandings.
Thinking before typing. "Limitation messages" encourage people to pause before posting something harsh. Time limits and taking turns lead to more thoughtful discussions.
Staying engaged. Tracking progress and positive feedback motivate people to keep participating.
What I would do next
Let people switch between the full chat and the mood view, make the AI summaries clearer, and use colour to show the tone at a glance.
Skills shown
Problem framingLiterature reviewUser interviewsEmpathy mappingExperiment designCo-creationAffinity mappingWizard of Oz testingDesigning AI featuresWorking with a client
Ana-Raluca Căruntu · UX Designer
I start with research, then design with precision.
UX designer with a Master in Digital Design and a BA in User Experience Design. I am looking for a UX/UI or product design role where research drives decisions.
Based in the Netherlands · Available now · No visa needed
Interviews, surveys and testing that lead to real design changes, in every project.
Designing for new kinds of interfaces
Apple Watch, haptic wearables and AI features. I learn the platform's rules first.
Keeping a team moving
Task tracking, workshops and playtests that took a team from first idea to a finished game.
Education
2024–25
Master Digital DesignHogeschool van Amsterdam
2021–24
BA User Experience DesignDe Haagse Hogeschool · minor Game Development Production
Experience
2024
UX graduation internship, Civic TechnologyDen Haag · AI moderation for Publieke Dialogen
2024
UX internship, NgraneAmsterdam
Working with AI
I use generative AI as part of my design process: to synthesise research, explore ideas faster, write and test prompts, and turn designs into working prototypes. I designed and built this website myself with AI-assisted development in Claude. My thesis was about designing AI features that people can trust.
Generative AIAI-assisted designAI-assisted prototypingPrompt designDesigning AI featuresClaudeChatGPTGeminiMidjourney
Methods
User interviewsSurveysPersonasCo-creationUser flowsWireframingPrototypingUsability testingWizard of Oz
