This project explored whether auditory cues could change how stiff a virtual object feels when pressed. The goal wasn’t to simulate stiffness with hardware, but to shift perception through multisensory experience. It challenged a core UX question: can we design for sensation without physical fidelity?

While haptic realism in VR often depends on expensive actuators, I was curious whether softness or hardness could be 'felt' differently using only sound. The challenge was to create a controlled experiment that stripped away variables and tested a simple idea: do brief, well-timed sound bursts alter our sense of material resistance?

Tellme began with a UX research question: how might we help researchers test and iterate surveys more efficiently, especially when they can't access target users? Inspired by the growing capabilities of LLMs, we hypothesized that we could simulate personas to generate realistic, structured responses to real-world surveys.

Through interviews with researchers, we surfaced key pain points: pre-testing surveys is slow, unrepresentative, and often biased. Users lacked tools to preview how different populations might interpret questions. Our insight was that what researchers needed wasn’t just better survey tools—but a way to simulate diverse user mindsets without conducting full deployments.

To evaluate the system’s usefulness and credibility, we replicated published survey studies using LLM-generated data. In multiple trials, synthetic responses produced similar distribution trends and latent structures to human data—giving us confidence in the platform’s viability for exploratory and prototyping use cases.

Set in a speculative 2035 Munich where a surreal 'Whispering Mist' distorts perception, Might Tower was conceived as a networked infrastructure system that helps cyclists navigate through reduced visibility and psychological disorientation.

The Whispering Mist affects visibility, cognition, and perception. Cyclists are especially vulnerable, struggling to stay visible to other vehicles and maintain spatial orientation. Interviews with both frequent and occasional cyclists uncovered fear and anxiety while navigating mist-dense areas. Traditional gestures and communication break down, and cyclists often push their bikes instead of riding.

Cycles is not a utility—it’s a temporal companion. Designed as a physical interface for slow living, it challenges our obsession with precision and speed by revealing time through imperceptible change. Instead of displaying numbers or blinking reminders, it changes shape so slowly that you never see it move—but always find it different when you return.

Modern life floods us with visual urgency: countdowns, alerts, blinking LEDs. But real time doesn’t behave like that—it flows silently, without notice. We asked: what if a device didn’t try to demand your attention, but instead existed quietly alongside you? Could its movement mirror the invisibility of time itself?

Cycles isn’t meant to be interacted with actively. In fact, we designed it so that you don’t notice it changing while you’re looking at it. But when you come back—from a break, a meeting, a walk—you realize something is different. That’s the experience of time. The lamp becomes a quiet signal that life is moving forward, even if we’re not counting it.

ACS360 was originally developed to help consulting teams visualize project health, forecast workload capacity, and make allocation decisions. However, the system had become outdated—slow to load, difficult to navigate, and hard to trust. I joined the project to lead the front-end redesign and improve how information was structured, visualized, and interpreted.

We began with contextual inquiries and stakeholder interviews. Consultants reported that they were relying on external spreadsheets because ACS360’s dashboard was too slow and cryptic to use in time-sensitive situations. They struggled to interpret raw tabular data, lacked comparative views over time, and found it difficult to identify which projects were at risk. Our problem became clear: the system had data—but it wasn’t providing insight.

Overwatch reimagines time tracking as a system that spans physical and digital space. Instead of forcing users to recall and re-enter activities, it captures context through a network of soft interactions: calendar sync, room sensing, and brief, intentional UI input. It’s a minimal interface layered onto an ambient system—designed to fit into how people already move through work.

Many knowledge workers—especially those balancing multiple projects, clients, or teams—struggle with fragmented time tracking tools. Retrospective reporting is cognitively demanding, and duplicative processes across internal systems and external formats lead to frustration. We asked: what if time tracking wasn’t a task, but an ecosystem of subtle cues that worked together to support memory, not replace it?

At the height of the COVID-19 crisis, public dashboards were flooded with numbers—but often failed to tell a meaningful story. We designed the Pandemic Dashboard to transform raw case and death statistics into a humane, interactive experience that helped people not just understand the data, but engage with it emotionally and contextually.

The German public faced a flood of disjointed information from news articles, government briefings, and social media. We saw a clear UX gap: while the data was technically open, it was not legible to most people. The challenge was to make complex, temporal datasets interpretable—and to allow for exploration without overwhelming users with noise.

One of the biggest design challenges was helping users explore without becoming disoriented. We solved this by implementing fixed headers, subtle animations to signal change, and embedded tooltips that revealed context on demand. Performance optimization was also key: datasets were sliced by region and time so the system only rendered visible data, ensuring speed and responsiveness.

Touchy Bonsai is a data sculpture that materializes the emotional traces of interpersonal touch. It invites users to reflect on past interactions through an open-ended, haptic dialogue. Built as part of a research-through-design process, it challenges the idea that data must always be encoded, categorized, or interpreted via language.

During a period marked by physical distance and increased digital communication, we asked: what does it mean to 'store' a moment of touch? And how might such moments be revisited—not through screens, but through tangible, sensory interaction? We aimed to design a system that preserves and reflects emotional residue, without assigning it predefined meaning.

We conducted a pilot study with international students who shared the feeling of social distance and relational disconnection. After journaling five meaningful interactions, participants used the system to express and revisit those experiences. Rather than interpreting the responses as feedback, they treated the tree as a companion—one that responds, breathes, and remembers without judgment.

This project tackled the gap between powerful machine learning systems and user understanding. I explored how interactive visualization could support non-expert users in understanding and trusting model outputs—specifically during hyperparameter tuning for text classification tasks.

In real-world ML workflows, domain experts often have limited control or insight into the systems they rely on. This opacity undermines trust—especially when users can’t see how model configurations affect outcomes. The core question: how can we design visualization tools that help users reason about and feel confident in machine learning behavior?