Aidena

— AI STACK RECOMMENDATION

AI Clinical Trial Matching Platform

End-to-end system to parse EHR records, extract patient eligibility criteria, and match patients to clinical trials using AI with scalable data pipelines and observability.

Stays alive for 365 days after the last visit.

Healthcare

AI Clinical Trial Matching Platform

End-to-end system to parse EHR records, extract patient eligibility criteria, and match patients to clinical trials using AI with scalable data pipelines and observability.

high confidence

Core Stack ℹ︎

AWS Textract

Primary

Automatically extracts structured text, tables, and form fields from unstructured EHR PDFs and scanned documents at scale with high accuracy for medical records processing.

$1.50/1k pages

Claude Opus 4

Primary

Most capable model for complex medical eligibility criteria extraction, patient-trial matching logic, and nuanced clinical reasoning with low hallucination on sensitive healthcare data.

$15/1M tokens

Airbyte

Primary

Open-source data integration platform to ingest EHR data from multiple hospital systems, normalize patient records, and sync to centralized data warehouse for trial matching pipeline.

$0/month (self-hosted)

Complete the Stack ℹ︎

Dagster

Alternative

Data orchestration platform modeling eligibility checks and patient-trial matching as software-defined assets with built-in lineage, quality checks, and observability for reliable clinical workflows.

$0/month (open-source)

Arize Phoenix

Alternative

Open-source observability framework to trace LLM calls during eligibility extraction, monitor hallucinations in clinical reasoning, and evaluate matching accuracy for compliance and safety.

$0/month (self-hosted)

Chroma

Alternative

Lightweight vector database to store trial eligibility criteria embeddings and patient clinical summaries for semantic matching and fast retrieval of relevant trials per patient.

$0/month (self-hosted)

Getting started

  1. 1Deploy Airbyte to connect EHR systems (Epic, Cerner, etc.) and normalize patient records into a PostgreSQL data warehouse.
  2. 2Use AWS Textract to extract structured data from unstructured EHR PDFs and clinical notes.
  3. 3Build Dagster DAGs to orchestrate daily patient record ingestion, validation, and quality checks.
  4. 4Use Claude Opus via API to extract eligibility criteria from trial protocols and match patients against criteria using few-shot prompting.
  5. 5Store trial eligibility embeddings in Chroma for semantic similarity matching.
  6. 6Instrument all LLM calls with Arize Phoenix to monitor extraction quality, detect hallucinations, and track matching confidence scores.
  7. 7Expose matched results via REST API with audit logs for clinical review and HIPAA compliance.
Copy link to clipboard

What are you building?

Build your own AI stack →