Software product · AI-native · In active development

Population-level risk, read the way a clinician would ask for it.

We build a software platform that turns lab results, assessments, and care-pathway records into risk-stratified, explainable guidance — at the population level and the individual patient level. Built as a product from day one, for health systems to deploy and extend, not a one-off engagement.

Request a walkthrough How it's built

Thesis

A platform, not a project.

Most AI pilots in health systems end as a single dashboard built for a single department. We're building the opposite: one reasoning layer, trained on the clinical workflow rather than one department's spreadsheet, that every module below draws from. New data sources extend the platform instead of forking it.

Contents

Eight modules, one reasoning layer.

Each module below is independently useful and already functioning end-to-end against live data. None are mocked for demonstration purposes.

Population

Risk & trend synthesis

Surfaces risks, trends, and recommendations across a patient population from aggregate metrics and category-level breakdowns.
Lab data

HL7 result interpretation

Parses raw HL7 v2.x lab messages, flags abnormal values against reference ranges, and recommends alerts or pathway changes.
Assessment

Per-assessment clinical review

Reads a single assessment in context — remarks, prior results, category — and proposes risks, pathways, and alerts with reasoning attached.
Pathway

Care pathway analysis

Scores pathway suitability, tracks progress against expected trajectory, and suggests next interventions.
Alerts

Active alert review

Checks existing alerts against current assessments and pathways for relevance, redundancy, and coverage gaps.
Search

Natural-language patient search

Turns a plain-English query — "patients over 60 in diabetes care with abnormal labs" — into structured filters.
Assistant

Patient-grounded AI assistant

Answers free-text questions about a single patient, strictly from that patient's own record — no inference beyond what's documented.
Core

Patient, program & pathway management

The non-AI backbone every module above reads from and writes to — patients, enrollments, pathway history, structured assessments.

Architecture

AI-native, not AI-decorated.

Every module above is powered by a frontier large language model reasoning directly over structured clinical data — not a rules engine with an AI label on top. Outputs are schema-constrained and validated before they ever reach a clinician's screen.

8 / 8

modules running against live model calls, zero stubbed

100%

AI outputs schema-validated before storage or display

reasoning layer shared across all modules — no per-feature forks

Current inference runs on a hosted frontier model. The architecture is model-agnostic by design — we are actively evaluating self-hosted and GPU-accelerated inference paths for health systems whose compliance posture requires on-premises deployment.

Audience

Built for the people who carry the caseload.

Not a BI tool for administrators to look at once a quarter — a working surface for daily care decisions.

Clinicians

Fast, explainable read on a single patient's trajectory — not a black-box score.

Care coordinators

A working list of who needs attention today, and why, instead of a static roster.

Population health teams

Cohort-level trend and risk visibility without hand-building a report each time.