UX Researcher & Product Designer
Uncover. Connect. Design.
I uncover root causes, connect perspectives, and translate research into clear product direction, helping teams build products grounded in evidence rather than assumption.
About
How I untangle a messy problem
What are people actually trying to accomplish?
How can complexity become something people can confidently use?
Skills
Research that uncovers patterns, clarifies uncertainty, and helps teams decide what to investigate or change.
Where I've Worked
Contributing to enterprise UX research across product teams, from research synthesis and UX measurement to AI-supported research workflows.
Worked across the UX research process, supporting user studies, research synthesis, product evaluation, and the translation of findings into actionable product insights.
Led cross-functional product development for automotive components, aligning design, engineering, and production teams from concept to manufacturing.
Designed automotive mechanical components from concept through technical documentation, collaborating with engineers and production teams throughout.
A selection of projects where research, systems thinking, and design come together to solve real problems.
Contact
I'm currently seeking full-time UX Research and Product Design opportunities in Germany and across Europe. If you're building products that solve complex problems through thoughtful research and human-centered design, let's talk.
Framework design, systems thinking, and end-to-end product design.
The problemEvery team measured UX differently, making results difficult to compare or use across products.
Designed a reusable UX measurement framework that integrates behavioral metrics, survey data, and expert evaluations into a shared decision-support system for enterprise product teams.
The problemInternational students arrive to a system that assumes they already know how it works.
Designed an end-to-end mobile experience helping international students navigate their first weeks in Germany, grounded in 4 qualitative interviews and 38 survey responses and validated through usability testing.
Synthesis, usability research, AI evaluation, and Human-AI trust studies.
The problemResearch existed across the organization, but the evidence was too fragmented to guide product decisions.
Synthesized interviews, prototype evaluations, support conversations, and user feedback into a shared evidence model, revealing recurring UX patterns, product risks, and actionable opportunity areas.
The problemA Microsoft Copilot Agent that sounded confident even when it was wrong.
Designed and evaluated a Microsoft Copilot Agent for reliable, source-grounded access to internal UX research knowledge.
The problemSmall friction points scattered across a grocery app were quietly costing user trust.
Conducted moderated usability testing with 8 participants using think-aloud protocol, identifying critical breakdowns in list creation, scan feedback, search relevance, and language accessibility, followed by targeted redesign proposals.
The problemWhen people cannot judge why an AI made a decision, knowing when to trust it becomes difficult.
Designed and ran a controlled, between-subjects experiment for IconChat, investigating how explainable AI reasoning shapes user trust, perceived usefulness, and intention to use for an AI-enabled app project.
Understanding the Existing System
The organization did not have a UX data problem. It had an interpretation problem.
Product teams already worked with established UX evidence, including survey metrics, behavioral metrics, and expert evaluations. These signals were valuable and were intended to remain unchanged: each metric already had its own purpose, calculation method, benchmark logic, and detailed scorecard.
Defining the Framework Opportunity
The framework was designed as a shared layer above existing UX metrics, not as a replacement for them.
The concept introduced one reusable interpretation system that connects existing survey, behavioral, and expert evidence while preserving the detail and context of each source.
The framework adds a structured summary layer that helps stakeholders understand:
The detailed scorecards remain the primary evidence layer. The unified framework acts as the interpretation and decision-support layer above them.
The framework connects fragmented UX evidence without compressing every signal into one oversimplified score.
Framework Evolution
The concept evolved through three stages.
The initial focus was understanding the existing measurement ecosystem. It then shifted toward connecting evidence through a shared interpretation layer, before finally defining the information hierarchy required for stakeholder decision-making.
Understanding where interpretation became difficult
The first sketch mapped the current UX measurement landscape.
Survey metrics, behavioral metrics, and expert evaluations already existed, but they were presented separately. The exercise made the central problem visible: stakeholders had to manually connect multiple signals before reaching a conclusion.
Mapping separate evidence sources, manual interpretation effort, missing confidence visibility, and limited trend comparison.
Connecting evidence without replacing it
The second sketch explored how existing UX signals could feed into one overall interpretation layer.
The concept introduced an overall UX summary, confidence visibility, comparison insights, conflict detection, AI-supported interpretation, and access to detailed evidence.
This was the point where the project shifted from designing a scorecard to defining a reusable evaluation framework.
Exploring how survey, behavioral, and expert signals could contribute to one transparent interpretation system.
Structuring the framework around stakeholder questions
The third sketch translated the framework concept into an initial information hierarchy.
The layout prioritized the questions stakeholders needed to answer first:
Testing the hierarchy of summary, confidence, evidence sources, comparison, and detailed interpretation before visual design.
Framework Specification
Before any interface was designed, the evaluation framework itself had to be defined. A detailed specification established how existing UX evidence should be aggregated, interpreted, compared, and communicated consistently across products. This document became the foundation for every design and implementation decision that followed.
From Framework Logic to Interaction
Once the framework logic was defined, I translated it into an interaction model through parallel sketching and low-fidelity exploration.
The goal was to test how stakeholders could move from an overall UX assessment to confidence, conflicting signals, metric-level evidence, and comparison over time without losing context.
These explorations helped define the page hierarchy, expandable scorecards, evidence relationships, and progressive disclosure before the concept was developed into an interactive prototype.
Rapid Interactive Prototype
The low-fidelity interaction model was developed into a complete interactive prototype to test the framework as a connected experience rather than as a set of isolated screens.
The prototype was first generated rapidly to accelerate exploration, then reviewed and refined in Figma. I adjusted the information hierarchy, interaction states, visual structure, and component behavior so the result reflected the framework requirements rather than the limitations of the initial AI-generated version.
It was used to evaluate:
The refined prototype became the shared reference for stakeholder discussions and engineering implementation.
Overview: overall UX grade, AI summary, and data confidence at a glance.
The refined interactive prototype became the foundation for implementation.
Throughout development, I worked closely with the developer to review and improve interaction details, information hierarchy, edge cases, confidence behavior, comparison logic, and component consistency.
The implemented experience continued to evolve through iterative design reviews and refinement during development.
Design Principles
Project Outcomes
Rather than introducing new UX metrics, the framework changes how existing survey, behavioral, and expert evidence is structured, connected, interpreted, and communicated.
It provides one shared model for understanding UX quality while preserving the context, confidence, and detail behind each conclusion.
Reflection
The project began as a challenge around presenting UX metrics more clearly. As the work progressed, it became clear that the deeper problem was not visualization, but interpretation.
The organization already had established survey metrics, behavioral evidence, expert evaluations, and detailed scorecards. What was missing was a shared way to connect these signals, communicate confidence, preserve contradictions, and compare UX quality consistently over time.
The most important design challenge was balancing simplicity with transparency. The framework needed to support quick stakeholder understanding without hiding evidence, uncertainty, conflicting results, or data limitations.
This shifted the work from designing a dashboard to defining a reusable evaluation system, including its interpretation model, information architecture, interaction behavior, confidence logic, comparison rules, and implementation specifications.
An AI research assistant is only as useful as the evidence behind its answers. I evaluated a Microsoft Copilot Agent across realistic research scenarios and turned what I found into concrete design decisions.
Due to confidentiality, internal prompts, proprietary research content, product-specific examples, and company data have been generalized.
The Reliability Problem
The agent could generate a fluent, confident-sounding response even when the underlying evidence was incomplete, weak, or missing. Failure could happen at more than one point along the way.
How I Evaluated Reliability
Evaluation Dimensions
Every response was checked against the same six dimensions, so reliability was judged as a whole interaction, not a single pass or fail.
One Evaluation Example
A representative, abstracted example. This is the same check I ran on every scenario.
Reliability Patterns
Sources were cited without being tightly connected to the specific claim they were meant to support.
Someone might trust a claim that only loosely traces back to its source.
Map the source hierarchy onto individual claims, not just the response as a whole.
When evidence was thin, the agent sometimes filled the gap with a plausible-sounding guess.
Confident synthesis and confident guessing read the same way to a stakeholder.
Constrain unsupported conclusions and flag responses where evidence is thin.
Knowledge boundaries weren't always stated explicitly, even across similar scenarios.
Unpredictable behavior undermines trust, especially in the cases where the honest answer is "I don't know."
Make explicit knowledge-boundary acknowledgment a required, consistent behavior.
Finding, Risk, Design Decision
Knowledge-Boundary Behavior
This wasn't proven empirically superior in a formal study. It's the reliability principle the evaluation pointed to, and the one the refined prompt architecture was built to enforce.
What Changed
My Contribution
What This Project Demonstrates
The hardest part was balancing usefulness with epistemic honesty: the agent needed concise, actionable answers that stayed traceable to evidence and honest about what the repository couldn't support. This project showed me how to evaluate AI systems beyond surface-level response quality, design for traceability, spot Human-AI reliability risks, and turn that evaluation into real design decisions. It also became the seed for my academic thesis, on how people perceive an AI system that's honest about the limits of its own knowledge.
Detailed Academic Report
The full report documents the research context, evaluation method, scenario design, prompt iterations, findings, and limitations in greater detail.
Read the Full Academic Report
Competitive Analysis
I explored existing apps and services for international students and newcomers in Germany to uncover gaps in usability, tone, and feature focus, and to define what an effective, student-centered onboarding experience could look like.
Free, official, comprehensive, suitable for newcomers.
Fragmented across regions, mostly in German, complex UX.
A student-specific experience with checklists for registration, bank account, insurance.
Comprehensive financial and insurance services, easier digital setup.
Narrow focus on legal setup, commercial rather than supportive tone.
Expand beyond finance into a friendlier, student-oriented tone.
Official and directly connected to universities.
Fragmented across regions, mostly German, complex experience.
Unified, simplified, translated information for international students.
Outcome Benchmark
I evaluated how well current apps deliver key outcomes international students need (clarity, guidance, momentum). Instead of listing features, I measured coverage and quality for each outcome.
| Outcome / Need | Ankommen | Expatrio | Studierendenwerk |
|---|---|---|---|
| Multilingual guidance | Good Clear multilingual content in English, Arabic, and others |
Partial English only |
Partial English only |
| Guided flow that prevents bottlenecks | None Static information |
None Focuses on financial products |
None Fragmented info by topic |
| Student-specific focus | None General newcomer guidance |
Partial Finance only |
Partial Fragmented, covers housing and enrollment in some regions but inconsistent overall |
| Localized, trustworthy instructions (city-specific, up-to-date) | None Provides general national info |
None National-level info |
Partial Local only, each local Studierendenwerk publishes separate info; varies in quality and accuracy |
| Actionable reminders & follow-ups | None | None | None |
| Guidance on what’s next (sense of completion) | None | None | None |
Survey Results & Insights
The survey collected both quantitative ratings and open feedback from international students. Below is a selection of the main findings, visualized to highlight the most critical challenges, sources of information, and feature priorities.
User Interviews
To complement the survey data, I conducted four semi-structured interviews with recently arrived international students. These conversations revealed emotional and practical struggles that numbers alone couldn't show.
Students don't just need a checklist. They need contextual guidance: required documents, the order of dependencies, and what happens if a step is missed.
Market Gap
Seven takeaways from the review shaped where Startklar's onboarding experience could set itself apart. Tap to read them.
User Persona & Stories
Interview and survey patterns converged into one primary persona, used to keep every design decision grounded in a real, specific person rather than an abstract user.
Job Stories
Framed as Jobs-to-be-Done: Sara's situations, motivations, and the outcomes the product needs to deliver for her.
Empathy Map
Translating interview and survey data into what Sara sees, does, hears, thinks, and feels, to separate her surface frustrations from what she actually needed.
Customer Journey Map
Mapping Sara's journey from preparing to leave her home country to settling into studies revealed how interconnected each bureaucratic step is. A single delay, such as waiting for an appointment, cascades through all following steps, amplifying stress and uncertainty.
Experience Mapping
Sara's experience shows a critical dependency chain: when one step is delayed, all following steps collapse. This domino effect leaves students stuck, anxious, and unable to proceed with their daily life.
The storyboard visualizes Sara's journey, from hopeful arrival to feeling stuck in paperwork.
Delays and unclear rules create a domino effect. Students feel stressed and isolated when they cannot progress independently.
Final Synthesis
Students don't fail because they are unmotivated, but because the bureaucratic steps are highly interdependent. Missing one step creates a chain of consequences that blocks others.
Students rely on fragmented sources like Telegram groups, informal advice, or vague official instructions. This patchwork of information fuels confusion and repeated mistakes.
Beyond everyday confusion, some students face extra barriers like sanctions, cash-only workarounds, and bank rejections, making already fragile timelines even more stressful.
Students arrive hopeful but quickly feel isolated, anxious, and unsupported. The uncertainty drains their focus and energy for academic life.
The issue isn't motivation; it's the absence of a clear roadmap. Students need a dependable, step-by-step guide rather than scattered encouragement.
Each city and university follows its own procedures. What works in one place can easily fail elsewhere, leaving newcomers lost between systems.
Since students were already overloaded with information, each iteration removed complexity instead of adding it. The goal became the lightest possible interface, so users could act without having to think.
Translating research insights into a clear product structure and two complementary navigation paths.
Information Architecture
Startklar sequences bureaucratic tasks by real dependencies and gives a browsable, trusted knowledge base, so students always know the next step.
To address the tension between sequential bureaucratic steps and the need for quick reference, I designed a hybrid navigation model.
A topic library for browsing by category (Banking, Housing, SIM, Community). Each topic page may reference related guided steps, but it never replaces them, so users can explore freely without losing the structured flow.
A linear timeline that keeps dependent tasks in the correct order. Later steps remain locked until prerequisites are complete. Each guided step opens a structured micro-flow with documents, appointments, and completion tracking.
After signing in, students choose between Guide Mode and Explore Mode, depending on how far they are in their onboarding journey.
Students indicate which steps they've already completed. The timeline automatically jumps to their current position, while later steps stay visible but locked. Each guided step opens a structured micro-flow with documents, appointments, and completion tracking.
A topic library for browsing by category (Banking, Housing, SIM, Community). Each topic page may reference related guided steps, but it never replaces them, ensuring users can explore freely without losing the structured flow.
This flow reduces confusion and prevents missed bureaucratic steps by locking later tasks until prerequisites are completed.
The Bank Account flow demonstrates the app's logic and interaction consistency.
Once a step is completed, it unlocks the next one in the timeline, creating a sense of progress and control.
Hand Sketches
Low-Fidelity Wireframes
High-Fidelity Prototype
Welcome & Onboarding Screens
Key App Screens
Bank Account Setup Flow
Usability & Iteration
Due to time constraints, I couldn't run full usability tests with real newcomers. Instead, I applied three complementary evaluation methods to uncover usability issues early and validate core interactions:
Heuristic Evaluation
Process: I created a matrix mapping Nielsen's heuristics against key screens and interactions. Each flow was reviewed step by step, friction points were noted and rated by severity, then clustered into themes like missing feedback or inconsistent labels. Findings were validated with two peers to confirm recurrent issues before iteration.
| Heuristic Principle | Observation | Severity | Action Taken |
|---|---|---|---|
| Visibility of System Status | Users couldn't tell when a sub-step was completed | Moderate | Added progress indicators and checkmarks after each task |
| Match Between System and the Real World | Bureaucratic terms (e.g., “Anmeldung”) were unclear for non-German users | Moderate | Added short plain-language explanations and translations |
| User Control & Freedom | Users couldn't easily edit or go back to previous inputs | Minor | Added a “Back” option and editable notes before submission |
| Consistency & Standards | Icons and navigation labels didn't match across flows | Minor | Unified icon-label pairs and adjusted hierarchy |
| Error Prevention | Users could skip dependent steps (e.g., opening a bank account before registration) | Critical | Locked later steps until prerequisites were completed |
| Recognition Rather Than Recall | Key navigation items were hidden inside menus | Moderate | Moved essential actions to the main navigation bar for better visibility |
The heuristic findings above formed the basis for the next validation phase: a cognitive walkthrough focused on first-time user behavior.
Cognitive Walkthrough – Simulating First-Time User Behavior
To simulate first-time user behavior, I performed a guided walkthrough of the prototype's key tasks (e.g., completing the “Bank Account Setup” step). Each subtask was assessed using three diagnostic questions:
Notes were attached as Figma comments, turning each friction point into a direct design annotation. This made it easy to visualize where clarity or feedback was missing inside the flow.
Peer Review & Mini Usability Tests
I invited three peers with similar relocation backgrounds to complete assigned tasks (e.g., “find bank setup information”). Each session followed a think-aloud approach, allowing me to observe real-time confusion points and compare them with heuristic notes. While limited in scope, these mini tests confirmed that the added feedback and simplified flows significantly improved comprehension and control.
Example Observation (from Evaluation Notes)
Visibility of System Status
Design Iterations Based on Findings
The evaluations directly led to several design improvements aimed at reducing friction and improving flow clarity:
Replaced redundant pages with modals to reduce context switching.
Added quick access from profile for frequent steps.
Added checkmarks and micro-confirmations after each task to reinforce confidence.
Integrated “Profile” in bottom navigation for predictable reachability.
Updated icon labels for better comprehension and consistency.
Planned Usability Metrics
While full user testing is planned for the next iteration, early heuristic and peer evaluations already indicated clear improvements:
These early metrics suggest that simplifying flow and adding micro-confirmations effectively reduce confusion and hesitation.
Accessibility Review
With limited time, I conducted a light accessibility audit focusing on color contrast, text readability, icon clarity, and keyboard navigation. Even these quick checks helped identify low-contrast elements and improved overall inclusiveness.
Outcome
The project evolved from an abstract idea into a structured onboarding system that provides clear direction when newcomers need it most. It demonstrates how clarity and small confirmations can reduce cognitive load and build emotional confidence during bureaucratic tasks. Designing and iterating in Figma helped transform a concept about “reducing chaos” into a tangible product ready for validation with real users.
Reflection
This project taught me to analyze user flows as attentively as emotions behind them. I learned that UX isn't just about clarity, it's about reducing stress through guidance and reassurance. Turning bureaucratic confusion into calm made me realize that empathy and structure can coexist, usability and emotional trust are equally important.
EDEKA is one of Germany's largest grocery retailers. Its app is used daily by millions, yet usability testing revealed a persistent gap between what users expected and what actually happened, particularly during list creation, product scanning, and purchase planning.
Usability issues in consumer apps are often trust issues in disguise. When users receive no feedback, they don't just get frustrated. They start to question whether the app is working at all.
Behind the scenes of a moderated think-aloud session: scenario script, live screen recording, and participant device.
Introduce a list overview as the default entry point, allowing users to immediately view all existing lists instead of landing directly on a previously used one. This surfaces the multi-list feature and lowers cognitive load during task execution.
Redesign the scanning experience to provide clear guidance and continuous feedback throughout the interaction. Show active scanning state, confirm when an item has been successfully added, and build confidence in the reliability of the feature.
Integrate price information directly into the shopping list experience, allowing users to view individual item prices while building their list and understand their overall spending. This enables informed decisions during the planning phase.
The most important insight from this project: usability issues in consumer apps are frequently trust issues in disguise. When the system provides no feedback, users don't just feel uncertain about the feature. They begin to doubt the entire product.
This project reinforced that feedback loops are not a nice-to-have. They are the foundation of a trustworthy experience. Every interaction without feedback is a missed opportunity to build user confidence.
A controlled, between-subjects experiment testing whether explainable AI reasoning shapes user trust, perceived usefulness, and intention to use, within a broader program of empirical Human-AI Interaction research.
Context
This research originated from a real product question brought to our university team by IconChat. For an AI-enabled app project, the company needed evidence on how explainability might influence users' trust and evaluation of AI. Our team designed a controlled experiment to investigate that question and provide research evidence that could inform the product direction.
"The experiment did not test whether people liked explanations. It tested whether explanations changed what people were willing to do with the answer."
Selected findings are presented here at a generalized level. Company-specific project information and non-public details have been omitted.
Study Design
Research Process
Key Findings
Research Decisions
Research Contributions
Reflection
Trust in AI is not built by accuracy alone. It is built by whether people can see enough of a system's reasoning to judge whether the output deserves belief. As AI systems take on more decision-support roles, that judgment becomes harder to make and more consequential to get right.
Four independent studies, four different formats. I built the shared evidence model that let recurring UX problems become visible for the first time.
Interviews, prototype tests, support conversations, and in-product feedback each had something real to say, just not at the same depth or reliability. Every observation kept its original source attached the whole way through.
How I Worked
A clear path from raw research to product direction.
From Evidence to Pattern
Rather than listing many disconnected findings, here's the full chain for one recurring issue. The same hierarchy applied across all five themes.
Patterns That Emerged
These mattered more than any single observation, because the friction showed up in more than one task, feature, or study.
What Changed
Useful findings, distributed across separate reports, questions, and formats. No way to see what recurred.
Recurring issues identified, patterns compared across studies, one coherent product-level view.
The synthesis also:
Product Implications
Keeping these three distinct was the point: it's what makes the synthesis traceable rather than a set of opinions. To get from pattern to implication, I framed each theme as a job the user was trying to get done, not just a complaint, so the recommendation stayed tied to what people actually needed rather than what broke on screen. These are research-grounded implications, not measured outcomes.
Broader consequences followed the same shape across themes: more learning effort, continued support dependency, and slower adoption of advanced capabilities. None of this was independently verified against business metrics.
Two Levels of Reading
Recurring themes, implications, and recommended areas of attention. Read in under a minute.
Observations, study context, and source relationships behind each theme. For scrutiny, not speed.
Reflection
Synthesis isn't summarizing several reports. It's preserving context, comparing evidence across sources, and making interpretation transparent enough to trust. The hardest part was balancing simplification with traceability. Some of the highest-value research work begins after the individual studies are complete.