"The right dose, every time"
AI-powered medication safety system that prevents errors in elderly patients through continuous vision monitoring, natural language interaction, and automated medication reconciliation.
125,000 Americans die annually from medication errors
- 50% of elderly patients don't take medications correctly
- Average elderly patient manages 10-15 daily medications
- Post-hospital discharge: 80% error rate in first week
- $300 billion in preventable hospitalizations
Current solutions (pill organizers, apps) only remind—they don't prevent.
TrueDose AI is an AI medication guardian that:
- 👁️ Watches medication cabinet 24/7 using Claude Vision
- 🛡️ Prevents errors BEFORE they happen (not just reminders)
- 🤖 Understands natural language conversations
- 📱 Alerts family in real-time
- 🏥 Reconciles medications after hospital discharge
Unlike existing pill dispensers that just beep on schedule, TrueDose AI SEES what medication is actually taken and prevents errors in real-time using Claude Sonnet 4.5's vision capabilities.
Watch this 2-minute video: Demo Link (Upload hackathon demo here)
Or run it yourself:
git clone https://github.com/yourusername/truedose-ai.git
cd truedose-ai
./scripts/quick_demo.shPatient reaches for wrong bottle (discontinued Aspirin)
↓
AI detects in 2 seconds
↓
"STOP! That medication was discontinued.
The correct medication is Eliquis (pink bottle on right)"
↓
Family alert sent
↓
Life saved
After hospital discharge, patients come home with:
- 6+ new prescriptions
- Changed dosages
- Discontinued medications
- Massive confusion
TrueDose AI:
- Scans all bottles with Claude Vision
- Identifies what changed
- Creates new schedule
- Flags dangerous conflicts
- Shares with family
Nobody else does this.
Patient: "Did I take my morning pills?"
TrueDose: "No, not yet. You usually take them at 8:15,
and it's 8:10 now. Your pills are ready."
Patient: "What's this white pill for?"
TrueDose: "That's Metformin for your diabetes.
You've been taking it for 3 years. It's safe."
- Detects confusion (opening/closing bottles repeatedly)
- Tracks unusual patterns
- Early dementia indicators
- Alerts family proactively
┌─────────────────────────────────────────────┐
│ TrueDose AI System │
├─────────────────────────────────────────────┤
│ │
│ [Camera] → [Claude Vision] │
│ ↓ │
│ Medication Analysis │
│ ↓ │
│ ┌─────────────┴────────────┐ │
│ ↓ ↓ │
│ Error Prevention Reconciliation │
│ ↓ ↓ │
│ Patient Alert Family Dashboard │
│ │
└─────────────────────────────────────────────┘
AI/ML:
- Claude Sonnet 4.5 (multi-modal: vision + language)
- Custom behavioral pattern learning
- Anomaly detection algorithms
Robotics:
- NVIDIA Isaac Sim (simulation & future dispensing)
- Real-time vision processing
- Precise medication tracking
Hardware:
- Raspberry Pi 5 (edge computing)
- 12MP wide-angle camera
- 4-mic far-field array
- Cost: $200 to build
Cloud:
- AWS (RDS, S3, EC2)
- Real-time family dashboard
- HIPAA-compliant data storage
- Ubuntu 22.04+ or macOS
- NVIDIA GPU (for Isaac Sim)
- Python 3.10+
- Anthropic API key
# 1. Clone repository
git clone https://github.com/yourusername/truedose-ai.git
cd truedose-ai
# 2. Run setup script
./scripts/setup.sh
# 3. Configure API key
cp .env.example .env
nano .env # Add your ANTHROPIC_API_KEY
# 4. Run demo
python src/main.py --demoFull installation guide: INSTALLATION.md
- 55 million Americans 65+ (growing to 80M by 2040)
- $455 billion home healthcare market
- Growing 12% annually
Consumer Direct:
- Hardware: $499 (or $99 subsidized with contract)
- Subscription: $49/month
- LTV: $3,000+ per customer
- CAC: $200 (LTV:CAC = 15:1)
B2B Partnerships:
- Assisted living facilities: 30,000 facilities
- Insurance companies: Medicare Advantage
- Clear ROI: 1 prevented hospitalization = $15,000 saved
5-Year Projection:
- Year 1: $793K revenue
- Year 3: $54M revenue
- Year 5: $380M revenue
Full business model: docs/BUSINESS_MODEL.md
| Feature | Pill Dispensers | Reminder Apps | TrueDose AI |
|---|---|---|---|
| Visual verification | ❌ | ❌ | ✅ |
| Error prevention | ❌ | ❌ | ✅ |
| Medication reconciliation | ❌ | ❌ | ✅ |
| Natural language | ❌ | ❌ | ✅ |
| Behavioral learning | ❌ | ❌ | ✅ |
| Works for dementia | ❌ | ❌ | ✅ |
- Technology: Multi-modal AI (Claude Vision) - cutting edge
- Data: Proprietary medication database + behavioral patterns
- Regulatory: Not a medical device - faster to market
- Network effects: More users = smarter system
truedose-ai/
├── src/ # Source code
│ ├── main.py # Main application
│ ├── claude_vision_guardian.py
│ ├── reconciliation_engine.py
│ ├── medication_manager.py
│ └── ...
├── docs/ # Documentation
│ ├── BUSINESS_MODEL.md # Complete business plan
│ ├── API_DOCUMENTATION.md
│ └── USER_GUIDE.md
├── hardware/ # Hardware specs
│ ├── HARDWARE_SPECS.md # Complete BOM
│ └── ASSEMBLY_GUIDE.md
├── pitch/ # Pitch materials
│ ├── PITCH_DECK.md
│ └── DEMO_SCRIPT.md
├── scripts/ # Utility scripts
├── tests/ # Test suite
└── README.md # This file
- ✅ Working prototype
- ✅ Hackathon demo
- 100-user pilot program
- Clinical validation study
- FDA clearance (if needed)
- 1,000 paying customers
- Insurance partnership (first)
- Series A fundraising
- 10,000+ customers
- National availability
- Assisted living rollout
- Platform expansion
- 1 million households using TrueDose AI
- Eliminate preventable medication deaths
- Standard of care for medication management
[Your Name] - CEO/Co-founder
- Personal connection: Grandmother's medication error led to hip fracture
- Background: [Your background]
[Co-founder Name] - CTO/Co-founder
- Built Isaac Sim integration
- Background: [Their background]
Advisor: Dr. [Name] - Geriatrician
- Clinical validation
- [Institution]
Seeking: $1-2M Seed Round
Use of Funds:
- 40% Product development
- 25% Hardware manufacturing
- 15% Clinical pilots
- 15% Marketing
- 5% Team expansion
18-Month Milestones:
- 10,000 customers
- $5M ARR
- Series A ready
Contact: hello@truedose.ai
- Installation Guide - Complete setup instructions
- Hardware Specs - BOM and build guide
- Business Model - Complete financial model
- Pitch Deck - Investor presentation
- API Docs - Developer reference
# Run test suite
pytest tests/
# Run specific test
pytest tests/test_vision.py -v
# Run with coverage
pytest --cov=src tests/We welcome contributions! See CONTRIBUTING.md for guidelines.
Areas we need help:
- Mobile app development (React Native)
- Hardware design (industrial design)
- Clinical validation (healthcare professionals)
- Beta testing (elderly users + families)
MIT License - see LICENSE file
- NVIDIA - Isaac Sim platform
- Anthropic - Claude Sonnet 4.5 API
- THE DROIDS STRIKE BACK - Hackathon organizers
- Beta testers - Early feedback and stories
Website: www.truedose.ai
Email: hello@truedose.ai
Twitter: @TrueDoseAI
LinkedIn: TrueDose AI
Hackathon Demo: Devpost Link
THE DROIDS STRIKE BACK - November 22, 2024
Categories:
- Best Use of Claude API
- Best Use of NVIDIA Isaac Sim
- Best Social Impact
- Overall Winner
Demo Video: YouTube Link
Pitch Deck: View PDF
"My grandmother was 78 when she came home from the hospital with 6 new medications. She got confused, took her OLD heart medication instead of the new one, fainted, fell, and broke her hip. Three weeks in the hospital. $45,000 in medical bills. This happens to 125,000 Americans every year.
We built TrueDose AI so no family has to go through this. The AI would have caught her error in 2 seconds. It would have alerted my mom. My grandmother would still be walking.
This is why we're here. This is why we're building. Join us."
— [Your Name], CEO & Co-founder
Last Updated: November 22, 2024
Version: 1.0.0
Status: Hackathon Prototype → Pilot Ready