TickDose | Medication, Voice-First

Voice-powered medication reminders with AI precision. Never miss a dose, stay on track, live healthier.

Comment

🎯 Inspiration: The 50% Problem

50% of patients don't take their medications as prescribed. This isn't laziness—it's forgetfulness, complexity, and lack of support.

Working with elderly patients managing 5+ medications, caregivers with no visibility into adherence, and frustrated doctors, we asked: What if medication reminders could speak? What if AI understood symptoms? What if family could help without being intrusive?

Voice + AI + Smart Tracking = Medication Adherence.

That's TICKDOSE.

🏗️ How We Built It: Enterprise Architecture

We didn't build a reminder app. We built a complete healthcare platform with 40+ microservices.

Technology Stack

Frontend:

  • Flutter 3.0+ (Android, iOS, Web)
  • Riverpod state management (40+ reactive providers)
  • Material Design 3 (dark mode, accessibility-first)
  • Multi-language support (English + Arabic with RTL)

Backend:

  • Firebase Authentication (email, Google, Apple, biometric)
  • Cloud Firestore (real-time sync + offline persistence)
  • Cloud Storage, Cloud Functions, FCM, Remote Config
  • Complete Firebase security rules (user data isolation)

AI & Voice:

  • Google Gemini AI (symptom analysis, medicine enrichment)
  • ElevenLabs text-to-speech (personalized voice reminders)
  • Google ML Kit (OCR medicine labels, barcode scanning)
  • Speech-to-text (hands-free voice input)

Local Storage:

  • Hive database (fast offline sync)
  • flutter_secure_storage (encrypted credentials)
  • SharedPreferences (settings)

12+ Core Features

  1. Smart Medicine Management - Manual entry, OCR, barcode scanning, AI enrichment
  2. Advanced Reminders - 8+ frequencies, timezone-aware, voice-enabled, quiet hours
  3. Medication Tracking - Calendar, daily timeline, adherence %, charts, PDF export
  4. AI Health Assistant - "I Feel" feature with Gemini for symptom analysis
  5. Voice-First Experience - ElevenLabs voice reminders + speech recognition
  6. Side Effects Logging - Track severity, timeline, patterns
  7. Caregiver Ecosystem - QR invitations, real-time alerts, adherence dashboards
  8. Pharmacy Finder - GPS-based, maps integration, offline data
  9. Offline-First Architecture - Full functionality without internet
  10. Multi-Language - English & Arabic with professional localization
  11. Accessibility - Biometric auth, voice commands, dark mode
  12. Security - HIPAA-ready, Firestore security rules, encrypted storage

40+ Production Microservices

Authentication (3): FirebaseAuth, BiometricAuth, PermissionService
Medicine (5): MedicineService, SideEffects, Barcode, Camera, OCR
Reminders (5): ReminderService, Notifications, FCM, VoiceReminders, TimezoneMonitor
AI & Voice (3): GeminiService, ElevenLabsService, SpeechService
Tracking (3): TrackingService, Analytics, PDFExport
Caregiver (3): CaregiverService, Sharing, Notifications
Utilities (15+): Cache, RateLimiter, RemoteConfig, Accessibility, DeepLinks, and more...

Smart Features in Action

OCR Medicine Capture:

  • Take photo of medicine label
  • Extract name, strength, dosage automatically
  • Gemini AI enriches with side effects & interactions
  • Add complex medicine in ~60 seconds

Intelligent Reminders:

  • 8+ reminder frequencies (once daily → custom schedules)
  • Meal-based triggers (breakfast, lunch, dinner, bedtime)
  • Timezone-aware (auto-reschedule when traveling)
  • Quiet hours enforcement (default 10 PM - 8 AM)
  • Actionable notifications (Taken/Skipped/Snooze)

Voice Interactions:

  • Personalized voice reminders (ElevenLabs)
  • Natural language symptom input ("I feel dizzy")
  • AI voice responses with emotional tone
  • Multi-language voice support
  • Real-time speech recognition

Caregiver Visibility:

  • Family can monitor adherence without intrusiveness
  • QR code invitations with permission levels
  • Real-time alerts for concerning patterns
  • Adherence rates & trends dashboard
  • Voice messaging to patients

🚧 The Challenges We Overcame

Challenge 1: Real-Time Sync + Offline-First

Problem: Users need reminders offline, but caregivers need real-time alerts. These seem contradictory.

Solution:

  • Implemented Firestore offline persistence with custom sync queue
  • Built conflict resolution for offline edits (last-write-wins with user override)
  • Created TimezoneMonitorService to reschedule reminders after timezone changes
  • Designed versioning system to prevent stale data

Learning: Real-time sync isn't magic—it requires obsessing over edge cases.

Challenge 2: Voice Reliability at Scale

Problem: ElevenLabs API calls fail sometimes. Voice reminders can't fail silently.

Solution:

  • Built ApiRateLimiter with per-user quotas to stay within budget
  • Fallback to local TTS if ElevenLabs fails
  • Cached voice files locally to reduce API calls 90%
  • Designed VoiceReminderService with 3-tier retry logic
  • Added voice health checks in settings UI

Learning: External APIs require graceful degradation and offline fallbacks.

Challenge 3: OCR Accuracy on Medicine Labels

Problem: ML Kit OCR works ~70% reliably. Medicine labels have varying fonts, angles, conditions.

Solution:

  • Added manual verification after OCR extraction
  • Implemented confidence scoring (show user only if >85% sure)
  • Created medicine database lookup for fuzzy matching
  • Used Gemini to validate extracted data ("Does this dosage make sense?")
  • Designed intuitive correction UI

Learning: AI alone isn't enough—good UX means letting users correct mistakes.

Challenge 4: Timezone Edge Cases

Problem: User travels Cairo → New York. Do reminders shift? How do we reschedule 20 reminders?

Solution:

  • Built TimezoneMonitorService (detects device timezone changes)
  • Smart rescheduling (if user traveling, reminders move with them)
  • Manual override option (user can keep original times)
  • Background WorkManager task (continuous monitoring)
  • Added timezone selection in onboarding

Learning: Mobile apps must think global, not just code correct.

Challenge 5: Caregiver Privacy vs. Visibility

Problem: Caregivers need visibility without becoming Big Brother.

Solution:

  • Granular permissions (view-only vs. edit access)
  • Caregiver sees aggregated adherence, not exact times
  • Alerts only for concerning patterns (3+ missed doses)
  • User controls what data caregivers access
  • Explicit invitation-based relationships (no surprise access)

Learning: Healthcare requires privacy-first design thinking.

💡 What We Learned

Technical Insights

  1. Flutter + Firebase = Scalable - Hot reload + real-time sync = fast iteration on production features
  2. Riverpod > Bloc - Composable providers beat state tree complexity every time
  3. Offline-First Unlocks Everything - Designing offline first made sync, error handling, and retry logic easier
  4. API Rate Limiting Saves Lives - Early designs called Gemini/ElevenLabs too much. Limiting + caching saved our budget
  5. Real-Time Listeners Are Expensive - Learned when to use streams vs. one-time reads

Product Insights

  1. Voice Changes Everything - Users ignore 10 text notifications but respond to 1 voice reminder
  2. Context > Frequency - Reminder at breakfast > 6 reminders at random times
  3. Caregivers Are Power Users - Expected daily patient use; discovered caregivers check constantly
  4. Design for Elderly = Good for Everyone - Large text, voice control, dark mode became beloved by all users
  5. Integration Unlocks Value - Pharmacy data, doctor contacts, caregiver workflows = 10x more value

Business Insights

  1. Problem Market Fit > Product Market Fit - Medication adherence problem so real, customers pay for partial solutions
  2. Healthcare Users Prioritize Privacy - Data security concerns dominate all other considerations
  3. Multi-Stakeholder Products Are Harder - Design for patients, caregivers, doctors = 3x complexity
  4. Voice AI is Still Emerging - Built for flexibility, not specific versions of Gemini/ElevenLabs

🏆 What Makes TICKDOSE Different

vs. Generic Reminders (Google Calendar, Apple Reminders)

  • ✅ Voice reminders (not just notifications)
  • ✅ Medication-specific logic (dosage, interactions)
  • ✅ AI health assistant
  • ✅ Caregiver support
  • ✅ Adherence tracking

vs. Existing Pill Organizers (Medisafe, Pill Reminder)

  • ✅ Conversational AI (Gemini)
  • ✅ Voice-first (most apps are text/tap)
  • ✅ Caregiver ecosystem
  • ✅ Production-ready architecture (40+ services)
  • ✅ Offline-first from ground up

vs. AI Health Apps (Ada, Babylon)

  • ✅ Linked to real medication management
  • ✅ Voice I/O for both sides (user asks, AI speaks)
  • ✅ Adherence focus (not just diagnosis)
  • ✅ Works offline
  • ✅ Family collaboration

📊 The Numbers

Scale:

  • 40+ microservices, 2 AI platforms, 3 platforms (Android/iOS/Web), 2 languages
  • 100+ data models with strict typing, 40+ UI screens, real-time sync + offline

Architecture:

  • Riverpod reactive providers, Firebase security rules, 20+ error types, rate limiting, background tasks

Features:

  • 12+ core features, 8+ reminder frequencies, 4 auth methods, OCR + barcode scanning, multi-language voice

🎯 Honest Challenges Remaining

  1. Healthcare Regulations - HIPAA/GDPR compliance designed in but not fully certified
  2. Doctor Integration - PDF export works; EHR integration (HL7, FHIR) needs partnerships
  3. Wearable Integration - Designed but not yet implemented (Apple Watch, Wear OS)
  4. Insurance Reimbursement - Requires clinical validation studies (in progress)
  5. Multi-Caregiver Support - One patient → one caregiver works; need permission trees for family
  6. Medication Database - Using Gemini for interactions; need production database for enterprise

🚀 What's Next

  • HIPAA Certification (Q1 2026)
  • Apple Watch & Wear OS (Q1 2026)
  • Pharmacy Integration (read prescriptions directly)
  • Multi-Caregiver Support (family councils)
  • Healthcare Provider Dashboard (doctors see adherence)
  • Clinical Validation Studies
  • 10+ Language Support
  • Insurance Partnerships

Stack Summary

Frontend: Flutter 3.0+, Riverpod, Material Design 3, Intl
Backend: Firebase (Auth, Firestore, Storage, Functions, FCM, Remote Config)
AI: Google Gemini (gemini-2.5-flash, gemini 3), ElevenLabs TTS, ML Kit, speech_to_text
Database: Firestore (cloud), Hive (local), flutter_secure_storage
State: Riverpod 3.0.3, Provider 6.1.1
Localization: Flutter Intl, full RTL support

🙏 Final Word

Building TICKDOSE taught us that the best technology is one people actually use. Medication adherence isn't solved by perfect algorithms—it's solved by a voice that cares, reminders at the right moment, and family watching over you.

If our app helps one person remember their heart medication and avoid a hospital visit—we succeeded. 💚

Built to solve medication adherence, one voice reminder at a time.

Built With

  • elevenlabs-tts
  • fcm
  • firebase-(auth
  • firestore
  • firestore-(cloud)
  • flutter-3.0+
  • flutter-intl
  • flutter-secure-storage
  • full
  • functions
  • gemini
  • gemini-2.5
  • gemini-3
  • hive-(local)
  • material-design-3
  • ml-kit
  • provider
  • remote
  • riverpod
  • riverpod-3.0.3
  • rtl
  • storage
Share this project:

Updates