Operationalizing Lifespan Wellbeing Science with AI
Built to support the mission of the WELLab at Washington University in St. Louis, this platform transforms wellbeing research domains into executable AI systems that accelerate discovery, intervention, and public impact.
Every WELLab research domain becomes a live AI module that:
- Captures data (EMA, wearables, clinical, self-report)
- Models dynamics (emotion coupling, causal inference, trajectory clustering)
- Generates insights (risk scores, protective factors, intervention targets)
- Deploys interventions (personalized coaching, activity prompts)
- Feeds back into research (continuous learning loop)
| Module | Research Domain | Key Capability |
|---|---|---|
| Real-Time Emotional Dynamics Engine | Short-term wellbeing fluctuations | EMA, IDELS emotion-coupling, volatility scoring |
| Behavioral + Physiological Health Engine | Wellbeing ↔ physical health | Causal inference (DoWhy), longitudinal regression |
| Lifespan Trajectory Engine | Long-term wellbeing change | Growth curves, trajectory archetypes, cross-cultural comparison |
| Cognitive Health & Dementia Prevention Engine | Wellbeing ↔ cognition / ADRD | Survival analysis, risk stratification, protective factor identification |
- Participant Experience — "Your Wellbeing Today", trend patterns, strength-framed insights (mobile-first)
- Researcher Dashboard — Coupling heatmaps, trajectory clusters, causal DAGs, data quality monitors
- Policy Dashboard — Population wellbeing maps, dementia risk distribution, intervention ROI (k-anonymized)
| Layer | Technology |
|---|---|
| Frontend | React, Vite, TypeScript, Tailwind CSS, Recharts, D3 |
| Backend | Node.js, Express, TypeScript |
| AI/ML | Python (scikit-learn, statsmodels, DoWhy, PyTorch), Anthropic Claude API |
| Infrastructure | AWS (Lambda, DynamoDB, S3, API Gateway, SageMaker, Cognito) |
| IaC | AWS CDK (TypeScript) |
| CI/CD | GitHub Actions |
wellab-platform/
├── SKILL.md # AI skill routing hub
├── README.md # This file
├── references/
│ ├── modules.md # Full specs for 4 AI modules
│ ├── data-model.md # Unified data model + DynamoDB design
│ ├── ai-capabilities.md # IDELS, temporal dynamics, bidirectional models
│ ├── dashboards.md # Participant, researcher, policy UIs
│ ├── architecture.md # API, infra, security, deployment
│ ├── ethics.md # Fairness, consent, scientific integrity
│ └── roadmap.md # Wearables, coaching agents, future phases
├── scripts/
│ └── fairness_audit.py # Demographic parity + disparate impact checker
└── assets/ # Templates, icons, fonts (future)
- IRB compliance — All participant data under approved protocol
- HIPAA-adjacent — Encryption at rest + in transit, audit logging, minimum necessary access
- Reproducibility — Pinned dependencies, deterministic seeds, version-controlled pipelines
- Cross-cultural fairness — Models audited for demographic bias before deployment
- Privacy — Individual risk scores never surfaced to unauthorized viewers; population data aggregated to k ≥ 10
# Clone
git clone <repo-url> wellab-platform
cd wellab-platform
# Install
npm install # Frontend + backend
pip install -r requirements.txt # ML pipelines
# Develop
npm run dev # Vite dev server
npm run api:dev # Express API (nodemon)
# Test
npm test # Jest + React Testing Library
pytest tests/ # ML pipeline tests
# Deploy
npm run deploy:staging
npm run deploy:prod # Requires PI + admin approval| Environment | Branch | Auto-deploy | Approval Required |
|---|---|---|---|
| dev | feature/* |
Yes | None |
| staging | develop |
Yes | None |
| production | main |
No | PI + admin |
This platform is built on WELLab's core principles:
- Reproducible AI pipelines with full audit trails
- Transparent model assumptions and confidence intervals
- Cross-cultural fairness audits (pre-deployment + monthly)
- Individual vs population safeguards at every layer
- Participant autonomy: view, export, or delete data at any time
See references/ethics.md for full details.
- Phase 1 (Active) — Core platform: 4 AI modules, data model, 3 dashboards
- Phase 2 (Planning) — Wearable integration (Apple HealthKit, Fitbit)
- Phase 3 (Research) — AI coaching agents (purpose, emotion regulation, social connection)
- Phase 4 (Concept) — Cognitive resilience training modules
- Phase 5 (Vision) — National wellbeing surveillance + clinical trial automation
Proprietary — Washington University in St. Louis. All rights reserved.
WELLab — Washington University in St. Louis