Thyroid care doesn't stop between appointments
Thyro is a clinical-grade mobile application that gives patients and care teams a continuous view of thyroid health. Lab results, symptoms, medication adherence, and biometric signals converge into a single longitudinal record — structured for pattern detection, not just logging.
The system is designed for long-term conditions where the clinical signal accumulates gradually and becomes meaningful only in context. Thyroid disease is the first application; the architecture is built to generalize.
Between visits, an assistant grounded in that record retrieves the right fragments at query time: labs, symptoms, adherence notes, and curated endocrinology references. Vector search keeps answers tied to the person’s own timeline rather than generic advice.
The product interprets, explains, and contextualizes. It does not diagnose or recommend treatment — those boundaries are enforced in prompts and architecture, not only in the interface.
Passive Apple Health signals (for example resting heart rate, heart rate variability, and sleep) enrich the same longitudinal view without extra manual logging, so patterns can be read in context over months and years.
RAG architecture
Vector search
LLM
Apple Health
Time-series data
Digital health
Good afternoon,
Marina
How are you feeling today?
Today's Logs
Medication
All doses taken
Heart Rate Range
Resting: 68 bpm
Range: 62 - 85 bpm
Sleep
7h 32m
Efficiency: 85%
Quick Actions
Symptoms & Medication
Track daily status
History
Review progress
Health Assistant
Ask your questions
Lab
Upload & analyze
What it is
Thyro is a clinical-grade mobile application for thyroid care: one structured, longitudinal record that spans labs, symptoms, medications, and biometrics — built so clinicians and patients can see how the story evolves, not only what happened on a single day.
The clinical gap
Most of the relevant data never reaches the consultation. Thyroid conditions evolve over months, not days. TSH, T3, and T4 fluctuate between quarterly lab draws. Symptoms surface and fade. Medication adherence drifts. These patterns accumulate in the interval between visits — and arrive at the consultation as fragmented recall, if at all. Thyro structures this interval data into a coherent clinical timeline. By the time a patient walks into an appointment, the relevant history is already organized, searchable, and ready to inform the conversation.
Pattern-based alerts for instability (in development)
Beyond tracking, Thyro is building a layer that surfaces early indicators of thyroid instability before they become clinically apparent. By analyzing longitudinal patterns across lab values, symptoms, and biometric signals, the system can flag combinations that may be compatible with an emerging flare, a T3/T4 imbalance, or a pre-symptomatic shift worth monitoring. This is not diagnosis. The system identifies pattern combinations that correlate with known instability profiles and surfaces them as signals for the patient and, where applicable, for clinical review. The threshold for alerting is conservative by design. Lab trend deviation: T3/T4 trajectories moving outside the patient's personal baseline, not just reference ranges. Symptom cluster onset: co-occurrence of fatigue, brain fog, and sleep disruption preceding or following lab changes. Biometric correlation: heart rate variability and sleep quality shifts that historically precede flare events in the user's own data.
Who it's for
Patients managing thyroid conditions — people living with Hashimoto's, Graves', hypothyroidism, or hyperthyroidism who want structured continuity between clinical visits and an early-warning layer when something starts to shift. Digital health teams and clinicians exploring AI-supported monitoring where longitudinal data interpretation adds value without displacing professional judgment — Thyro shows what that boundary looks like in practice.
History
Symptom
Medication
Levothyroxine
75 mcg • 08:00 AM
28/30 doses taken this month
Heart Rates
Monthly average
Heart Rate Range
Resting: 68 bpm
Range: 62 - 85 bpm
Sleep
Monthly average
Sleep
7h 32m
Efficiency: 85%
TSH
TSH Levels (mIU/L)
Last 18 months (tests every 3 months)
Trend: TSH levels have improved significantly over the past 18 months
What it tracks
Onboarding and consent-scoped context
From the first session, users define their clinical context and choose what categories of data to share. That consent-gated foundation keeps the timeline and assistant aligned with the person’s situation and privacy choices.
Welcome to Thyro
A quick setup so Thyro can personalize your endocrine health experience.
Select Condition
Assistant grounded in the patient's record
A conversational assistant uses RAG: each query pulls the most relevant fragments from the user’s own record — labs, symptoms, adherence — together with curated endocrinology documentation, so answers reflect real longitudinal data rather than generic medical information.
Health Assistant
Hello Marina, how are you feeling today?
Can you explain my latest TSH results?
Your latest TSH from January 15, 2024 is 3.1 mIU/L, which is within the normal range (0.4-4.0 mIU/L). Good news - this is actually lower than your August reading of 3.5, showing improvement. Your levels have been stable around 3.1-3.5 over the past few months, and you're doing well managing your Hashimoto's. Keep tracking and we can monitor any changes together!
Lab results
PDF and image upload with automated biomarker extraction. TSH, T3, T4, and custom markers are tracked longitudinally. Each result sits on the timeline and is compared against prior values, not only reference ranges.
Lab Results
Upload Lab Result
Select a PDF file
Your Lab Results
result-Nov-2024.pdf
PDF Document
result-Aug-2024.pdf
PDF Document
result-May-2024.pdf
PDF Document
result-Feb-2024.pdf
PDF Document
result-Nov-2023.pdf
PDF Document
result-Aug-2023.pdf
PDF Document
Biometric signals
Passive Apple Health integration for resting heart rate, heart rate variability, and sleep quality. These signals enrich pattern analysis without adding manual input friction.
Settings
Marina
Preferences
Customize what information you want to track
Weight Tracking
Track body weight
Daily Reminders
Receive a gentle nudge at 8:00 PM if you haven't logged your day.
Apple Health
Heart Rate and Sleep data are automatically synced from Apple Health
Privacy & Consent
Manage your privacy consents and data usage
I accept the Terms of Use and Privacy Policy
Required for full use of the app.
I consent to the processing of my health data
Includes symptoms, medications, lab results, and medical history.
I consent to the use of medical AI tools
Symptom logging and medication adherence
Daily structured input with free-text observations; patterns can be viewed across weeks and months, with frequency and co-occurrence analysis to surface recurring clusters. Dose-by-dose tracking for levothyroxine and other thyroid medications, correlated with symptom and lab timelines to spot gaps in the record.
Daily Log
Current Symptoms
Pattern-based alerts (in development)
A planned layer will highlight conservative, explainable combinations of labs, symptoms, and biometrics that may warrant attention — always as signals for discussion with a clinician, never as a standalone diagnosis.
Reports
Symptoms Summary
Last month
Total Symptoms Logged
24
Most Common
Heart Rate
Heart Rate Range
Resting: 68 bpm
Range: 62 - 85 bpm
Sleep
Sleep
7h 32m
Efficiency: 85%
Create Your Report
AI architecture: grounded in the patient's own data, bounded by clinical safety
The assistant uses a RAG architecture. At every query, vector search retrieves the most relevant fragments from the patient's own record — lab values, symptom history, adherence data — and combines them with curated endocrinology documentation. Responses are grounded in longitudinal data, not generic medical information.
Retrieval layer: vector search over personal health data dynamically pulls the right labs, symptom entries, and historical context at query time. Knowledge layer: clinical reference material is indexed alongside personal data so the assistant can contextualize findings within established thyroid physiology.
Safety boundary: no diagnosis, no treatment recommendations. The system interprets, explains, and contextualizes. It does not generate clinical decisions. This boundary is enforced at the prompt and architecture level, not just in the UI.
Security & privacy
Privacy model: user-owned, consent-gated data. Information is stored with explicit per-category consent. Users control what is shared, when, and with whom. The product is progressing through formal security and quality certification.
AI behavior stays constrained: no automated clinical decisions, clear separation between education and decision support, and engineering choices that reinforce the same safety posture end to end.
Let's talk
Building something similar for your patient population or clinical workflow? We can talk through the architecture, the safety model, or what it would take to adapt this for a different chronic condition.