‹ back to portfolio

Women's health intelligence

Magdalene X²

One picture of your whole health.

Connecting labs, cycle, sleep, nutrition, and fitness through plain-language insights and small personal experiments.

Hi-fi prototype · Wellness insight, not diagnosis
Role
Founder · Product & Design
Type
0→1 Product Design
Stage
Concept → Hi-Fi Prototype
Year
2026

Overview

Magdalene X² is an autonomous, multi-agent health-intelligence companion that unifies a woman's labs, menstrual cycle, nutrition, fitness, and sleep into one system. It detects meaningful changes, explains them in plain language, generates hypotheses, and runs small closed-loop personal experiments — framed as wellness insight, never diagnosis.

Every woman deserves a tireless health analyst that understands her whole biology — not five disconnected apps. How might we turn the data a woman already produces into understanding and action?
Product Design Multi-Agent AI Health Tech N-of-1 Experiments Femtech UX Research

Problem & Motivation

Health-engaged women generate more personal health data than ever — rings, watches, CGMs, lab panels, cycle apps — yet no tool connects it. They get numbers, not meaning, and they're left to play detective across five domains that biologically interact.

318K+
health apps exist with no synthesis layer
~6%
of sport-science studies are female-only
25%
more of their lives women spend in poor health vs. men
~18%
of self-trackers ever share data with a clinician

The menstrual cycle is a master modulator — it affects glucose response, sleep architecture, injury risk, and nutrition needs. Any single-domain reading (e.g., HRV or a lab value) is biologically ambiguous without cycle context. Yet no consumer tool reads these domains together.

User Research & Market Analysis

Conducted deep research across 30+ products in three categories — biomarker platforms, AI health assistants, and femtech — to map the competitive landscape and identify the unoccupied white space.

The Competitive Gap

Product Cycle Sleep Fitness Nutrition Labs Autonomous AI Self-Experiments
Flo (77M MAU) ✅⚠️❌❌❌⚠️❌
Oura ✅✅⚠️❌❌⚠️❌
WHOOP ⚠️✅✅❌⚠️⚠️❌
Levels ❌⚠️⚠️✅⚠️⚠️⚠️
Wild.AI ✅⚠️✅⚠️❌⚠️❌
Allara (clinical) ⚠️❌⚠️✅✅❌❌
Fountain Life ❌⚠️⚠️⚠️✅✅⚠️⚠️
Magdalene X² ✅✅✅✅✅✅✅
🔗
Cross-Domain Synthesis

No product integrates all five domains — labs, cycle, nutrition, fitness, sleep — into a unified analysis.

🤖
Autonomous Agents

"Agentic" in consumer health today means a chatbot that can call an API. No one deploys multi-agent orchestration.

🧫
Closed-Loop Experiments

No consumer tool automates the N-of-1 workflow: baseline → intervention → measure outcome → learn.

Target Users

Three personas form a clean, non-overlapping spectrum — prevention → performance → investigation — and sequence the system by complexity.

👩🏻‍💻
POC · Primary
Olivia — The Healthy Optimizer

Mid-20s, healthy and asymptomatic. Wears an Oura ring, logs workouts, curious about a CGM. Wants to prevent future problems and build durable habits.

"I have five apps and a spreadsheet and still can't tell whether what I'm doing is actually keeping me healthy."

🏃🏽‍♀️
MVP · Secondary
Mia — The Athlete

Late 20s, serious recreational athlete. Energy and recovery swing with her cycle but nothing explains it. Suspects low iron.

"Generic fitness AI tells me to push harder on days my body can't."

👩🏾‍⚕️
GA · Secondary
Priya — The Dismissed Patient

Mid-30s, experiencing irregular cycles and energy crashes. Wants answers and evidence to bring her clinician.

"Doctors say 'your labs are normal' and move on."

Key Scenarios

Olivia · Prevention
Sleep × Glucose × Cycle

Magdalene detects that short sleep raises next-morning glucose in the luteal phase → proposes a 2-week sleep-protection experiment → shows flattened spikes.

Mia · Performance
Recovery × Ferritin × Cycle

Connects poor late-luteal recovery + low-normal ferritin → recommends phase-shifted intensity + iron-focused fueling → re-checks and reports recovery change.

Priya · Investigation
Glucose × Cycles × Clinician Report

Connects prolonged post-meal glucose in the luteal phase + irregular cycles → produces a clinician-ready summary + a protein-first-breakfast experiment.

Design Process

I explored five visual directions by studying well-designed health and femtech apps, then converged on a direction that combines Headspace-style modularity (clear, grouped modules) with Co–Star-style personalization (a voice that knows you).

Phase 1 — Research
UI Case Study

Analyzed 12+ health and femtech apps across five visual directions (calm dark-premium, warm editorial, clinical data-forward, bold athletic, personality-driven) to identify patterns worth borrowing.

Phase 2 — Concept
Low-Fi Concept

Defined the design language — warm cream palette, editorial serif for the "voice," clean sans for data. Five core screens: Today, Modules, Insight, Ask, and Goals.

Phase 3 — Iteration
Design Iterations

Refined typography (Fraunces + Inter), interaction patterns, and the experiment flow through multiple rounds. Added interactive tab navigation and smooth transitions.

Phase 4 — Hi-Fi
Hi-Fi Prototype

Full-fidelity interactive prototype with all five screens, interactive elements, and the complete insight → experiment → outcome flow.

Key Design Patterns Borrowed

  • One number per domain first — never a table on the home screen (inspired by Oura, WHOOP)
  • Color = state, not alarm — green/amber mean "where you are," not "good/bad"
  • Plain-English "what this means" after any data point (Function, WHOOP)
  • Drill-down, not dump — full data is 2–3 taps deep, never the default
  • Deviation & rest = information, not failure — anti-guilt framing throughout

Prototype Iterations

Every module was designed in code, not in a static canvas. Each step below is one iteration: a real, working comparison build put side by side so the choice could be judged on the actual screen rather than a description of it. The five modules here took 41 iterations: color families, ring geometry, message tone, and data-entry placement, each one narrowing the next decision.

Took

Left

Why

System Architecture

The product loop: Ingest → Detect → Explain → Hypothesize → Experiment → Learn. The experiment engine writing outcomes back to the timeline is what closes the loop — the core differentiator.

📊
Ingest
Wearables, cycle, labs, CGM, nutrition → unified health timeline, cycle-phase tagged
→
🔍
Detect
Autonomous change & pattern detection across all domains — no user prompt required
→
🤖
Reason
Multi-agent orchestrator coordinates domain agents, hypothesis engine, experiment designer
→
🛡️
Guard
Safety layer screens outputs — no diagnostic claims, honest uncertainty, clinician escalation
→
💬
Surface
Plain-language insights, experiments, weekly summary, clinician report

UX Principles

💡
Say what it means, not how it works

Lead with the insight, hide the machinery. The user sees the meaning — not the multi-agent pipeline behind it.

🎯
One clear next action

Every insight offers at most one small experiment. Reduce decision fatigue to a single "yes."

⚖️
Honest uncertainty

Never over-claim. Show confidence levels, defer to clinicians, and always frame as "information, not a diagnosis."

🔒
Privacy you can feel

End-to-end encryption, no third-party data sharing, granular consent. Visible control over data as a trust signal — especially for reproductive data.

Product Design

The final design uses a warm, editorial voice paired with modular data cards, with the cycle phase treated as a first-class variable that contextualizes every insight. The live hi-fi prototype is embedded below — the design system first, then six screens. Onboarding, Modules, Experiments and Me are interactive: tap their cards and buttons to walk the nested pages.

Magdalene X² — hi-fi prototype open full size ↗

Outcome & Status

Magdalene X² targets a category-creation white space that the market is visibly self-organizing toward — but no incumbent has executed: cross-domain synthesis + autonomous agents + a closed experimentation loop.

The project produced a comprehensive product requirements document, deep competitive and scientific research, three validated user personas, a full UI case study across 12+ competitor apps, and an interactive hi-fi prototype — all the artifacts needed to move from concept to build.

Next Steps

  • POC (Olivia): Connect data + deliver one real cross-domain insight + one semi-manual experiment. Prove the "aha."
  • MVP (Mia): Cycle-aware training/fatigue insights; guided experiments; multi-source integration; retention instrumentation.
  • GA (Priya): Handle hormonal/metabolic complexity; clinician-ready reports; scaled autonomous experiment engine.