Khalisa by Evia Wellness — Cycle-Aware Hormone Intelligence Platform
Built a clinician-supervised platform that treats the female hormone cycle as a vital sign — a patent-pending hormonal digital twin that unifies wearables, symptoms, labs and medical records into one system, with Claude running on Amazon Bedrock inside Evia's own AWS account. Not a period tracker. A clinical intelligence platform.
Women's healthcare has a measurement problem. Heart rate, blood pressure, and glucose are all treated as vital signs — tracked continuously, interpreted clinically, acted on early. The female hormone cycle, which governs sleep, mood, metabolism, cardiovascular risk, and bone density across four decades of life, is treated as lifestyle data.
The consequence shows up most sharply in perimenopause. Only about one in four US women aged 45–64 who could benefit from menopause treatment actually receive it.[1] Untreated menopause symptoms cost the US an estimated $26.6 billion a year in lost work time and direct medical expenses.[2] Not because the therapy doesn't exist, but because nobody is measuring the signal that would indicate a need to change that. Evia Wellness set out to change that with Khalisa™ — a platform that models each woman's hormonal biology continuously, predicts where she is heading rather than reacting to where she has been, and puts a clinician in the loop on every decision. Bitsol built it.
Turning a cycle into a clinical signal
Hormone health generates enormous amounts of data across wearables, symptom logs, lab panels, and medical records — none of it structured, connected, or clinically actionable. Khalisa™ needed to unify all four into a system a clinician could safely prescribe from.
- The hormone cycle drives sleep, mood, metabolism, cardiovascular and bone health, yet sits outside standard clinical monitoring
- Care is reactive: women present with symptoms long after the underlying shift began
- Perimenopause is the sharpest failure point — women say they were misdiagnosed when they sought care, often treated for anxiety or depression while the hormonal shift went unaddressed
- Existing consumer apps predict periods, not disease pathways
- Continuous wearable streams — sleep, heart rate variability, body temperature — arriving in three different vendor formats
- Daily subjective symptom logs with no structure a model can learn from
- Quarterly lab panels in inconsistent units and reference ranges
- Medical records that clinicians trust but algorithms can't parse
- All of it PHI, all of it requiring HIPAA-grade handling from the first line of code
A hormonal digital twin, with a clinician in the loop
Computational modeling of each woman's reproductive lifespan, fed by a unified data layer, surfaced through a clinician portal with review and override at every step.
- Patent-pending clinical architecture: the hormonal model, the titration framework, and the clinician-in-the-loop design
- Maps each woman's position across her reproductive lifespan
- Predicts which disease pathways are emerging, rather than reporting what already happened
- Enables cycle-aware medication titration so therapy adapts to her biology
- Thirty days of forecast ahead, updated daily, on a calendar she can read
- Wearable integration across Apple Health, Oura and Fitbit
- Symptom and cycle tracking with pattern recognition
- Patients upload lab reports as they receive them; Claude reads each one so a result from one lab can be compared with a result from another
- Medical record ingestion into the same patient model
- Clinician portal with full review and override workflows
- Clinician-approved therapy schedules surfaced to the patient
- No autonomous clinical decision — a licensed clinician signs off
- HIPAA-compliant architecture, PHI encrypted at rest and in transit
- Built with guidance from experts in computational pharmacology, clinical pharmacy and women's health
Khalisa™ launched as a clinician-supervised beta in June 2026 and has been in production with patients in perimenopause since July 2026, where it has produced more than 1,100 forecast runs to date. The architecture extends across the full female lifespan — puberty, postpartum, and menopause.
Claude reads. The clinician decides.
Khalisa™ runs on Claude, in the live product. When a patient uploads a lab report, Claude reads it and pulls out the values that matter. When her wearable logs three weeks of broken sleep and rising overnight temperature, Claude reads that alongside her symptom history and her labs, and tells her what the next thirty days are likely to look like — day by day, symptom by symptom.
Then it does what a period tracker never does. It drafts the order. For a patient whose symptom pattern holds at the same severity through every phase of her cycle rather than spiking in the luteal week, Claude writes out its reasoning in full and hands it to her clinician: what it saw in the data, why that pattern points to a structural cause rather than a cyclical one, what it checked against the medications she already takes, and what it wants the clinician to confirm before anything is prescribed.
The clinician signs it, changes it, or throws it out — and writes down why, every time. That record cannot be edited afterwards. Nothing reaches the patient without it.
Behind that sits the engineering: the calculations run in ordinary code before any model is called, patient-written text is screened before it reaches Claude, every drafted medication is checked against two federal registries, and the platform measures its own forecasts against what patients actually record. Claude Haiku 4.5 handles the high-volume reading; Claude Sonnet 4.6 does the clinical reasoning. All of it runs on Amazon Bedrock inside Evia's own AWS account, under Evia's Business Associate Agreement with AWS — patient data never leaves the environment they control.
What changed
Khalisa™ moves women's healthcare from reactive to preventive. Rather than waiting for a woman to present with symptoms, the platform models her hormonal trajectory continuously and flags emerging pathways while there is still time to intervene.
The forecasts are being checked against reality. In its first nine weeks of production use, Khalisa™ generated 9,050 daily symptom predictions for an initial cohort of 20 patients. Patients confirmed 323 of those predictions as matching what they actually experienced, and rejected 53. Every confirmation writes a real symptom record; every confirmation and rejection feeds back into the accuracy of the next forecast.
For the patient.
She sees what is coming, not what already happened. Thirty days of forecast, on a calendar, in her own symptom vocabulary — and when the platform gets it wrong she can say so, and the next forecast accounts for it.
For the clinician.
A drafted order arrives with the reasoning attached: what the data showed, what was checked, what still needs confirming. The work of reconciling a wearable export, a symptom diary, a lab panel and a chart note is already done.
For Evia.
A clinician-supervised platform live with patients, built to extend from perimenopause across puberty, postpartum and menopause without rearchitecting.
Delivery: 8 weeks, concept to launch. Production figures cover 2026-06-24 to 2026-08-27, Khalisa's initial patient cohort.
[1] NORC at the University of Chicago and the AARP Public Policy Institute, analysis of Medical Expenditure Panel Survey data.
[2] Mayo Clinic, survey of 4,400 US women aged 45–60.
Straight From The Client
Reviewed at beta launch, June 2026
Bitsol helped us launch the beta. Our users started receiving daily AI symptom forecasts from cycle, wearable, and lab data right away, replacing guesswork with guided care.
Related Work
Want to launch your HIPAA-compliant MVP in 8 weeks?
Tell us your idea — even if it's still rough. We'll map your 8-week path in 30 minutes.