We are building the intelligence layer that turns dose data into clinical action. It is designed to identify patterns, surface emerging concerns, support timely outreach to patients, and give clinicians the context for the moments that matter, on a foundation proxy-based systems cannot offer.
In the design we are developing, the detection of a problem, the classification of the patient, and the selection of a response are rule-based and clinician-reviewed. The model personalizes the wording of a message rather than deciding the clinical action, and a clinician stays in the loop on every recommendation.
Its responses are being built to operate under the MyAide AI Constitution, drawing on three foundations: Anthropic’s constitutional AI methodology, the AMA Code of Medical Ethics, and a trustworthiness framework derived from the foundational consumer trust and accountability research of our founding team.
The system is being developed to use retrieval-augmented generation over a curated clinical knowledge base, with explicit retrieval confidence thresholds and a structured self-evaluation step before each response.
MyAide is being built to read each patient’s measured dosing behavior, combine it with a behavioral profile drawn from brief instruments grounded in our research, and recommend a specific, clinician-reviewed intervention matched to that patient. The behavioral model places each patient on the dimensions that predict response, and pairs that with the dose-level record only quantity measurement can produce. The result is a recommendation that reflects both what the patient did and why they are likely to respond to one approach rather than another.
