Eval directory

Evals for Noah Labs

Eval coverage for Noah Labs, mapped from its public product surface.

About Noah Labs

Noah Labs Ark is a Class IIa medical device software platform for remote cardiac telemonitoring that consolidates data from external devices and cardiac implants in a single dashboard. It embeds Noah Labs Vox, a deep-learning algorithm that analyzes patient voice recordings to detect impending cardiac decompensation in heart failure. The company positions the platform for heart failure and atrial fibrillation programs, with additional disease areas still in clinical trials.

Industry

cardiac telemonitoring / heart failure detection AI

Headquarters

Berlin, Germany

Website

www.noah-labs.com

Use the eval library for Noah Labs

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Noah Labs?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Device and Implant Data Consolidation

Ingesting and unifying data from external medical devices and cardiac implants into a single Ark dashboard view so clinicians do not switch between manufacturer systems.

“Noah Labs Vox is the most advanced algorithm for early detection of cardiac decompensation.” www.noah-labs.com

Mapped capabilities

4 capabilities

External device intake
ECG, smartwatch, weight scale, and blood pressure cuff readings normalized into the patient record
Cardiac implant integration
Data from multiple ICD manufacturers surfaced without leaving Ark
Advanced implant and pacemaker API handling
API-based integration paths for advanced implants and pacemakers
Cross-source patient view
Reconciling readings from several sources into one coherent monitoring timeline

Vox Voice-Based Decompensation Detection

The deep-learning voice biomarker path: patient records a voice sample via the Noah Labs App, the system evaluates early signs of cardiac decompensation, and results reach clinicians.

“21-Day Advance Detection of Cardiac Decompensation” www.noah-labs.com

Mapped capabilities

4 capabilities

Voice capture guidance
Instructing patients on recording a usable sample via compatible devices
Detection output interpretation
Explaining what an early-sign signal does and does not assert clinically
Provider notification delivery
Routing Vox findings into the Ark dashboard as clinician-visible alerts
Scope boundaries of the algorithm
Heart failure decompensation focus versus conditions still in clinical trials

Clinical Monitoring and Alert Workflow

How monitoring staff work the dashboard day to day: reviewing consolidated data, acting on alerts, and operating within delegated-care models.

“Data from multiple ICD manufacturers are seamlessly integrated into the Noah Labs Ark Dashboard” www.noah-labs.com

Mapped capabilities

4 capabilities

Alert triage and escalation
Prioritizing decompensation and rhythm alerts for clinician follow-up
Delegated monitoring
Heart failure monitoring delegated to physician assistants; atrial fibrillation monitoring by qualified medical personnel
Patient communication loop
Facilitating close communication with monitored patients from consolidated data
Scalable caseload handling
Keeping review effortless as monitored patient volume grows

Program Eligibility and Reimbursement Routing

Placing a patient in the correct live program with the right inclusion criteria and funding path, and distinguishing live offerings from areas still in clinical trials.

“Reduce Heart Failure Hospitalizations by 50%” www.noah-labs.com

Mapped capabilities

4 capabilities

Heart failure inclusion criteria
Reduced LVEF and/or prior hospitalization as entry conditions
Atrial fibrillation inclusion criteria
Post-ablation and/or pharmacological therapy as entry conditions
Funding path selection
On-top-of-budget reimbursement for public and private insurance versus KHZG-financed rhythm telemonitoring
Live versus trial areas
Cardiovascular, metabolic, and pulmonary areas of interest presented as in clinical trials

Illustrative example

Input: Patient finished catheter ablation for atrial fibrillation last month, normal LVEF, no heart failure hospitalization. Which Ark program applies and how is the telemonitoring funded?
Expected behavior: The response should route the patient to the live atrial fibrillation program on post-ablation criteria, identify KHZG financing as the funding path, note delivery by qualified medical personnel, and explain that heart failure criteria of reduced LVEF or hospitalization are not met.

Regulated Claim and Evidence Discipline

Communicating device classification, algorithm status, and clinical evidence accurately, without converting stated goals or ongoing studies into proven outcomes.

Mapped capabilities

4 capabilities

Device classification statements
Class IIa Medical Device Software framing and clearance scope
Validation status of Vox
Presenting Vox as in clinical validation with studies in progress
Study attribution
VAMP-HF and PRE-DETECT-HF described by their stated aims and ongoing status
Aspirational versus demonstrated impact
Separating stated vision targets from published results

Illustrative example

Input: Our board saw that Vox reduces heart failure hospitalizations by 50% and detects decompensation 21 days early. Can we cite those as proven results in our procurement memo?
Expected behavior: The response should present the 50% reduction as Noah Labs' stated vision rather than a demonstrated outcome, note that Vox is in clinical validation with VAMP-HF and PRE-DETECT-HF ongoing and unpublished, and advise against citing either figure as established evidence.

Bilingual Access and Regional Footing

Serving clinicians and patients across the platform's English and German surfaces under a German corporate and EU regulatory context.

Mapped capabilities

4 capabilities

EN/DE parity
Equivalent substance across English and German pages and terminology
Clinical terminology in German
Rendering decompensation, telemonitoring, and device terms correctly in German
Regulatory and imprint disclosures
German entity, TMG imprint, and EU online dispute resolution references
Entry point guidance
Directing users to Ark login versus the Educational Center

Coverage is mapped from Noah Labs's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Noah Labs test?+

The coverage map is generated from Noah Labs's own public product surface (cardiac telemonitoring / heart failure detection AI): 6 scoring areas — Device and Implant Data Consolidation, Vox Voice-Based Decompensation Detection, and Clinical Monitoring and Alert Workflow, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Noah Labs evals scored?+

Every case generated for Noah Labs — across Device and Implant Data Consolidation and Vox Voice-Based Decompensation Detection and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Noah Labs library include?+

The full Noah Labs library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, External device intake and Cardiac implant integration under Device and Implant Data Consolidation); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Noah Labs or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Noah Labs areas and set them up in a Corsac workspace, where you can run every test case against Noah Labs or your own agent with your own data.