What I build, how it works, and what I can prove
Designed and built end to end by Margarita Montañez Davenport — product, UX/UI, integration engineering, consent architecture, and deployment.
Google Health API readings mapped to FHIR R4 Observations with LOINC codes. In the live demo, a visitor’s wallet-signed, revocable consent grant issues 15-minute access tokens per metric, and the grant’s permission-tree root can be anchored on Ethereum Sepolia testnet. Sleep shows a dated Google Health API snapshot; heart rate and steps show labeled sample data.
Open the wearable consent demo →
The signed-consent module maps wearable readings from the Google Health API into FHIR R4 Observation resources with LOINC coding. Google OAuth 2.0 gives the module scoped, time-limited access to the account. The participant signs one consent grant covering heart rate, steps and sleep; it issues three 15-minute access tokens, one per metric, and revoking the grant ends all three at once. Each grant and revocation is signed by the participant’s wallet and kept as an entry in a permission tree in the participant’s browser; anchoring the tree’s root on Ethereum Sepolia testnet lets a disclosed entry be checked against it. Daily ingestion runs on the investigator’s own device; the public demo shows a dated snapshot.
Authentication and informed consent happen in a single cryptographic event: the person signs a human-readable consent statement with their wallet, and that signature simultaneously proves identity, records consent, and authorizes the session. No username, no password, no email. Permission scopes are explicit and enumerated. Revocation is available at any time; each single-grant revocation is a wallet signature recorded as a verifiable event, and revoking every grant at once runs from the signed-in session, without a separate signature. On encrypthealth.io, the sign-in message is a plain-language EIP-191 statement that carries the consent text and a single-use nonce, so signing in IS signing the consent text.
An internal security review identified findings across the deployed infrastructure. Remediation is verified finding by finding on the security page. The architecture operates with a zero-PHI data model: participant records are keyed to a pseudonymous wallet address, and the database schema has no fields for a participant’s name, email address, phone number, date of birth or IP address. Wallet addresses are persistent pseudonymous identifiers; the regulatory posture rests on operating as neither a covered entity nor a business associate. All ten current contracts (13 deployed, 2 superseded, 1 awaiting production switch) are source-verified on the Blockscout explorer.
I own the product end to end: discovery, requirements and acceptance criteria, architecture, roadmap, and release decisions. Delivery runs on AI-assisted, multi-agent pipelines under written scope contracts with build-artifact verification. Every feature ships against stated acceptance criteria. The platform — from the consent architecture to the FHIR mapping to the smart contracts — was designed, built, and deployed under my technical and strategic leadership.