Florian Schulte

Whistle needed a coaching layer that could do more than display workouts on a calendar. The Coach project explored how an AI assistant could understand intent, remember training context, and turn progress data into guidance that feels timely, personal, and easy to act on.

To create a truly personalized coaching experience, we needed to design a system that felt less like an app and more like a human partner. We focused on long-term memory as a core design pillar, allowing the coach to build a dynamic model of the user's goals and conversation history. This ensures every interaction feels continuous, relevant, and naturally guided toward their fitness journey.

We designed the daily and post-workout summaries to deliver high-level insights the moment a user opens the app. By generating these progressively throughout the day, we eliminated loading friction. Key data points are surfaced using a structured chip system, allowing users to instantly understand their progress without reading dense text.

To keep the experience engaging, the coach delivers playful, contextual notifications that celebrate accomplishments by anchoring current workouts against past performance, making every milestone feel meaningful.

I took ownership of this project from design through development and release. The work was end to end: running user interviews, designing the system, implementing it, and shipping it.

Outdoor workouts depend on conditions that change by the hour, but most weather interfaces ask athletes to interpret raw forecasts themselves. Weather for Whistle reframed the forecast around the actual training decision: when conditions are right, risky, or worth adjusting for.

When designing the weather experience for outdoor workouts, our goal was to reduce cognitive load by surfacing only the most actionable data. Instead of overwhelming users with raw meteorological data, we designed a high-density weather chart that allows athletes to assess conditions at a glance.

To make this data immediately relevant, we process hourly forecasts into clear, workout-specific summaries such as "Perfect Conditions," "High Wind," or "High UV." This allows users to instantly understand how the weather impacts their specific exercise without having to decipher raw numbers.

To further streamline the decision-making process, we introduced Best Times. By analyzing complex weather parameters against ideal workout conditions, we automatically rank and surface the optimal windows for exercise throughout the day. This shifts the user experience from passive data consumption to proactive, guided planning.

I took ownership of this project from design through development and release. The work was end to end: running user interviews, designing the system, implementing it, and shipping it.

Rewind was a hackathon prototype for Meta smart glasses that explored a provocative question: what if wearable capture could become a searchable memory layer? The project tested continuous visual indexing and multimodal recall while making the privacy tension impossible to ignore.

During the Start Hack Hackathon, we developed an application for Meta smart glasses that continuously captures the wearer's visual environment, utilizing an embedding model to seamlessly index and organize the data.

The application also features a live mode designed for next-generation augmented reality displays. By continuously processing real-time audio and video feeds, the system intelligently distinguishes between moments the user wants to save and active search queries. Users can simply ask, "Where did I leave my keys?" to instantly retrieve the relevant visual memory, or point at an object and verbally confirm they want to remember it.

I worked together with Riccardo Bussetti on this hackathon project. He took over the backend and I designed and built the iOS app. The project won the community award.

Training load can help athletes progress safely, but the raw metrics are hard to translate into everyday decisions. This Whistle project turned exertion, workout history, and near-term predictions into an interface that helps users understand when to push, adapt, or recover.

Designing the training load feature was centered on translating complex physiological data into actionable, user-friendly insights. The goal was to create an intuitive interface that empowers athletes to balance their exertion and progress safely, removing the guesswork from their daily training decisions.

To make the data truly valuable, we designed a predictive model that forecasts the next seven days of training load based on individual workout metrics like distance, time, calories, and effort. This shifts the user experience from simply reviewing past performance to actively planning a safe, sustainable progression.

I worked closely with my cofounder Clemens Scholz on this project. Clemens kicked off the project and did the initial research and backend implementation. After that, I took over to create the final interface.

Peaks Insights turned sleep, recovery, and overnight health signals into a clearer picture of day-to-day wellbeing. The project focused on making complex personal data understandable enough to guide behavior, not just report what happened while the user was asleep.

The Insights feature in Peaks was designed to transform abstract sleep data into a meaningful, holistic overview of the user's well-being. Rather than just displaying numbers, the interface focuses on actionable intelligence that helps users understand the interplay between rest, recovery, and daily life.

When designing a composite score, we always start with the core user goal: shifting behavior. For sleep, the objective was to help users build a stable circadian rhythm. We carefully weighted the underlying metrics to prioritize consistency, ensuring the interface naturally guides users toward healthier habits.

The Recovery Score visualizes how well-rested a user is in relation to their physical exertion. By connecting rest directly to exercise, the design helps athletes intuitively understand when to push harder and when to pull back.

Vitals provide a baseline for general wellness by tracking overnight physiological signals. The design focuses on subtle, early indicators, allowing users to spot potential illness before it fully manifests.

As a key driver of circadian health, regularity needed a prominent place in the app. We intentionally placed more weight on the wake-up time, the one variable users have the most direct control over, empowering them to take charge of their rhythm.

I worked closely with my cofounder Clemens on this project. He created the backend for vitals and the recovery score while I worked out the interface and the sleep and regularity scores. The work was designed on paper, then built using Swift and SwiftUI.

FeatherKit is the design system behind Vogelhaus apps. Version 5 focused on the shared foundations, component logic, surface styling, navigation, and shader effects that let multiple products feel consistent while still adapting to their own context.

At Vogelhaus we build multiple apps that share the same technologies. This includes RevenueKit for purchases, InsightsKit for tracking and A/B testing, and ServerKit for the backend. The kit we have been working on the longest is FeatherKit, the design system that powers our apps.

The most fundamental parts of FeatherKit are the colors and text styles. They integrate well into the system, and custom modifiers coordinate accent colors for components as well as color styles for surfaces. The context a component is shown in is passed by modifiers that change its appearance.

Components have their style and logic split. This makes it easy to validate the logic while creating multiple styles for the same component. A picker can come in various shapes and forms, including simple lists, cards, or circles for icon pickers. The logic is always the same; only the presentation changes.

Components are aware of each other. Lists in groups render differently, as do lists in buttons. Groups understand their nesting level and padding so the correct corner radius can be set automatically.

Since the introduction of shaders in SwiftUI, we are building a library that is used to add life into our apps in various ways. Shaders allow for more expressive effects, and you can find them throughout our apps.

The navigation stack coordinates sheets, banners, and alerts that sit higher in the view hierarchy.

Peaks centers on circadian rhythm: an ambient model of how energy changes across the day. This project translated that biological pattern into glanceable guidance across the app, widgets, and watch surfaces so users could plan around their rhythm without constantly opening the app.

The circadian rhythm serves as the foundational element of Peaks. It visualizes the user's daily energy fluctuations, transforming abstract biological data into actionable, glanceable guidance.

To make the data meaningful, the design surfaces Influences, the external factors like exercise, caffeine, or melatonin that shift the user's natural rhythm. By visualizing the cause-and-effect relationship of these actions, the interface empowers users to consciously shape their daily behavior. Optimal Times further contextualizes this data by highlighting ideal windows for focus, exercise, or winding down, allowing users to align their schedule with their biology.

Because the circadian rhythm is inherently passive and slow to shift, the core UX challenge was building awareness without creating friction. The design utilizes glanceable, low-footprint widgets across iOS and watchOS, including the Smart Stack, to normalize the user's biological states such as the afternoon dip, reframing them as natural rather than unproductive. By leveraging relevance APIs, the interface intelligently surfaces these widgets at the most impactful moments throughout the day, ensuring the rhythm remains a helpful, ambient guide without requiring active engagement.

I worked closely with my cofounder Clemens on this project. We ran many interviews to find the direction and designed the interface together. He created the backend for the rhythm calculations while I implemented the interface. The work was designed in Figma and built using Swift, SwiftUI, and WidgetKit.