Amorphous AI builds a data platform that converts unstructured hospital records into standardized, queryable clinical data and layers an AI reasoning engine on top.
Funding
Funding not disclosed
Founders
Product
Problem
Hospitals generate massive amounts of clinical data, but the majority is stored in unstructured formats such as PDFs, free‑text notes, and inconsistent EHR fields, making it inaccessible to automated analysis and decision‑making tools.
Solution
Amorphous AI provides a data platform that extracts information from unstructured hospital records and maps it to standard clinical ontologies (SNOMED, LOINC, ICD‑10, RxNorm). The structured data becomes queryable via a multi‑agent reasoning engine called Striata, which automatically generates reproducible SQL queries, Python analysis code, statistical tests, visualizations, and written reports. Each step includes provenance metadata, enabling auditable, minutes‑fast answers to population‑scale clinical and operational questions.
Target Audience
Primary customers are hospital data and analytics teams, clinical research departments, and health system operations groups that need rapid, reproducible insights from their existing clinical documentation.
Features
- AI‑driven extraction pipeline that converts free‑text notes, PDFs, and heterogeneous EHR fields into standardized entities with 97.8% recall.
- Mapping of extracted entities to major clinical ontologies (SNOMED CT, LOINC, ICD‑10, RxNorm) for interoperable data.
- Multi‑agent reasoning layer (Striata) that translates natural‑language questions into cohort definitions, SQL queries, statistical analyses, and visualizations.
- Automatic generation of audit artifacts: executable SQL, Python scripts, and interactive visualizations with full provenance tracking.
- Sub‑minute response latency (~2 minutes) for end‑to‑end query‑to‑report cycles.
- Support for custom statistical tests (e.g., Welch’s t‑test) and flexible report formatting for clinical and research audiences.