Peri

AI Recovery Intelligence for Perimenopausal Athletes

Brief

Peri is an AI-assisted health intelligence platform designed for active women in perimenopause. Rather than simply tracking symptoms, the product translates wearable signals, cycle data, training history, and self-reported symptoms into recommendations that help users understand what is most likely affecting their capability and recovery today.

Why Peri

Women in perimenopause often don't lack information.

They lack interpretation.

Most health apps focus on tracking symptoms or predicting cycles. But fluctuating hormones, changing recovery capacity and inconsistent responses make generic recommendations increasingly unreliable.

Peri emerged from a different question:

How do you restore trust in an unpredictable body?

My Role

Founder, Product Designer, UX Strategist and System Architect.

Peri became an exploration of AI-native product design and multi-model workflows. I directed product strategy, research synthesis, information architecture, interaction design and system behaviour while orchestrating AI tools across design and implementation.

What I did

Designed for uncertainty

Most health products focus on tracking individual metrics or presenting raw data. Peri was designed around a more difficult problem: helping users understand what is most likely affecting their capability and recovery on any given day.

To support this, I designed a recommendation framework that combines wearable data, cycle information, symptoms, training history, and recovery signals. The system adapts to both regular and irregular cycles, accounts for incomplete data, and prioritizes the strongest signals rather than relying on rigid rules. This allowed the experience to behave more like a reasoning system than a traditional tracker.

Built trust through explainability

Health recommendations are only useful if users understand and trust them. Rather than presenting opaque scores or unexplained advice, I designed a progressive disclosure framework that exposes the reasoning behind recommendations.

Users can move from a concise summary to increasingly detailed explanations, revealing which factors influenced the recommendation and how strongly they contributed. This approach helped balance simplicity with transparency and avoided the "black box" problem common in AI-driven experiences.

Created reusable interaction patterns

As the product evolved, maintaining consistency across increasingly complex features became critical. I established reusable interaction patterns and shared design principles that governed recommendations, trends, insights, confidence indicators, and empty states.

These patterns reduced complexity while ensuring that similar problems were solved consistently across the product. Rather than designing isolated screens, I focused on building a system of reusable behaviors that could scale as new capabilities were introduced.

Documented the product architecture

To support future development and AI-assisted workflows, I documented the product beyond traditional UI artifacts. The project includes feature inventories, state maps, information architecture, user flows, decision models, and a structured data model.

These artifacts capture not only what the interface looks like, but how the system behaves, how information flows, and how decisions are made. This level of documentation makes complex products easier to evolve and creates a shared understanding between design, product, and engineering disciplines.

Designed and built the product end-to-end

Peri was created as an AI-assisted solo project, spanning product strategy, research, interaction design, information architecture, design systems, documentation, and implementation.

The final product includes a fully functional Next.js application, wearable integrations, recommendation engines, synchronization workflows, and more than 250 automated tests. By combining design thinking with implementation, I was able to move rapidly between ideas, prototypes, and production while maintaining consistency across the entire experience.

Design principles

Design for uncertainty

Perimenopause rarely follows predictable patterns. Instead of relying on rigid rules, I designed the experience to adapt to incomplete and conflicting signals from wearables, symptoms, cycle history, and training. Recommendations are based on the strongest available evidence and adjust to both regular and irregular cycles.

Make recommendations explainable

Health insights are only useful if users trust them. I used progressive disclosure and confidence-aware language to reveal not just what the product recommends, but why. This approach avoids black-box behavior and helps users understand the factors influencing each recommendation.

Build systems, not screens

As the product grew, consistency became more important than individual interfaces. I established reusable patterns, shared behaviors, decision models, and documented architecture to ensure the experience could evolve without increasing complexity.

Results

Designed and built a fully functional AI-assisted product from concept to implementation.
Created a capability and recovery intelligence system that transforms wearable signals, symptoms, cycle data, and training history into actionable recommendations.
Established a complete architecture portal documenting feature inventories, state maps, user flows, decision models, and design principles.
Built a scalable system supported by 250+ automated tests and reusable interaction patterns.
Created foundations for future AI-driven health experiences that prioritize transparency, trust, and explainability.