Medical Validation & Clinical Standards for AI Blood Test Analysis
An internal, automated benchmark of the Kantesti AI engine, authored by our Chief Medical Officer and not independently validated or peer-reviewed, behind the AI Blood Test Analyzer.
Primary Evidence & Documentation
The performance figure on this page comes from our own automated, open-source benchmark harness, with code and scorecards published openly. It is a reproducible technical benchmark, not an independent or regulatory validation, and not a measure of real-world diagnostic accuracy. Access the code and data below.
Kantesti AI Engine — Blood Test Interpretation Benchmark
Automated composite scoring on 100,000 synthetic cases across 127 country labels
Benchmark & Methodology Overview
Kantesti's AI Blood Test Interpretation platform undergoes rigorous medical validation as part of our internal quality process. Performance is measured with an internal, automated, open-source benchmark harness.
This page documents that benchmark, our physician-oversight structure, and our continuous quality-monitoring process. The benchmark is a technical evaluation, not a clinical trial or regulatory validation.
Internal Benchmark Performance (Composite)
This is the result of our internal, automated, open-source benchmark harness. It is self-run and has not been independently validated or peer-reviewed.
Overall Benchmark Result
Composite benchmark score: 99.80% (self-run, not peer-reviewed) — the overall result of our automated, open-source benchmark harness on 100,000 synthetic cases across 127 country labels. The composite is a blended technical metric (35% output-structure validity + 55% clinical-keyword recall + 10% latency); it is not a measure of diagnostic accuracy, and it has not been independently validated or peer-reviewed. See code & data on GitHub
The benchmark produces a single composite score from an automated harness; it does not produce per-test-category sensitivity/specificity figures. Full per-case scorecards are in the open repository.
Medical Advisory Board
Our Medical Advisory Board provides clinical oversight for all AI algorithm development and validation. Board members represent diverse specialties across multiple countries, across clinical medicine.
Thomas Klein, MD
Chief Medical Officer (CMO) Clinical Hematology & AI DiagnosticsDr. Sarah Mitchell, MD, PhD
Chief Medical Advisor Clinical Pathology & Laboratory MedicineProf. Dr. Hans Weber
Senior Medical Advisor Laboratory Medicine & Clinical ChemistryDr. Maria Rodriguez, MD, MPH
Medical Advisor Internal Medicine & Preventive MedicineDr. Chen Wei, MD, MSc
Medical Advisor Endocrinology & Metabolic MedicineMeet our full Medical Advisory Board with detailed profiles, credentials, and research backgrounds.
View All Advisors →Continuous Quality Monitoring
Post-deployment, the engine is monitored through internal operational checks and structured feedback from users and our Medical Advisory Board. This is internal monitoring and is not an independent clinical-outcome study.
Monthly Performance Reports
Comprehensive accuracy analysis across all biomarker categories, demographic segments, and geographic regions. Trend identification enables proactive quality management.
Healthcare Provider Feedback
Structured feedback integration from physicians and laboratory professionals. Flagged interpretations undergo Medical Advisory Board review with corrections integrated into training.
Training Data & Quality Assurance
Our quality-assurance process applies standard controls to the data used by the platform.
Quality Controls
Multi-stage data quality assurance:
- Removal of incomplete or corrupted records
- Outlier detection for pre-analytical errors
- Provenance verification for all datasets
- HIPAA-aligned and GDPR-aligned anonymization
Technology & Compliance Partners
Our validation infrastructure and AI development is supported through partnerships with industry-leading technology providers.
Microsoft FoundersHub
Cloud infrastructure and enterprise-grade AI development platform supporting scalable validation workflows.
NVIDIA Inception Program
GPU computing resources and AI model optimization enabling efficient model training and optimization.
Google Cloud AI
Machine learning infrastructure supporting distributed model training and real-time inference.
Cloudflare
Global edge network ensuring secure, low-latency access globally.
HIPAA-aligned
US healthcare data protection safeguards
GDPR-aligned
European data protection regulation
Appropriate Use & Limitations
Transparency about capabilities and limitations is essential for responsible AI deployment in healthcare. Kantesti is designed as a decision support tool to complement—not replace—professional medical judgment.
Information Not Accessed
Our AI interprets biomarker data in isolation. The following clinical context is not available to the system:
- Complete patient medical history
- Current medications and potential interactions
- Physical examination findings
- Genetic factors and family history
- Lifestyle factors (unless user-provided)
Laboratory Methodology Variations
Reference ranges vary between laboratories due to equipment differences and calibration standards. Our database of laboratory-specific reference ranges addresses many variations, but users should verify extracted values match their original report.
Document Quality Considerations
OCR accuracy depends on document quality. Handwritten results or low-resolution scans may affect value extraction. Manual correction is available for all extracted values before analysis.
Medical Disclaimer
Kantesti is an AI-powered informational tool that interprets blood test results based on established medical reference ranges and clinical guidelines. It is NOT a medical device and does not diagnose, treat, cure, or prevent any disease.
Information provided is for educational and informational purposes only and should not be considered medical advice. Always consult with a qualified healthcare professional before making decisions about your health or treatment.
For medical emergencies, contact emergency services immediately. Kantesti is not designed for emergency situations.
Conflict of Interest & Funding Disclosure
This validation documentation is published by Kantesti Ltd (UK Companies House No. 17090423), headquartered in London, United Kingdom. Medical Advisory Board members receive compensation for their advisory roles. The Chief Medical Officer (Thomas Klein, MD) is a full-time employee of Kantesti Ltd. All validation reported here is internal to Kantesti and has not been independently verified or peer-reviewed. No external funding was received for validation studies. The company is self-funded through revenue and technology partnerships with Microsoft for Startups, NVIDIA Inception, Google Cloud, and Cloudflare.
Frequently Asked Questions About Medical Validation
How often is the AI model updated?
Our model undergoes quarterly retraining incorporating new validated data, updated clinical guidelines, and emerging biomarker research. Each update is checked against our internal automated benchmark before deployment. Updates that don't meet our benchmark threshold are not shipped.
How is benchmark performance reported?
Performance is reported as a single composite score from our automated benchmark, not as per-category diagnostic accuracy. The composite blends output-structure validity, clinical-keyword recall, and latency, and is a technical measure only.
Can I trust AI interpretation for medical decisions?
Kantesti is designed as a decision support tool, not a replacement for professional medical judgment. Our AI provides validated interpretations based on established reference ranges, but clinical context—including your medical history, medications, and symptoms—requires physician evaluation for treatment decisions. Always consult with your healthcare provider.
How many cases and country labels does the benchmark cover?
The automated benchmark runs on 100,000 synthetic cases spanning 127 country labels. It is a technical content-and-format benchmark, not a demographic accuracy study, and has not been independently validated or peer-reviewed.
What happens if the AI makes an error?
Healthcare providers and users can flag interpretations for Medical Advisory Board review. Flagged cases are analyzed by our CMO Thomas Klein, MD, and the medical team. If errors are confirmed, corrections are integrated into future training cycles. Our continuous monitoring tracks real-world performance to identify and address systematic issues proactively.
Where can I find the full validation report?
Our benchmark methodology, code, and per-case scorecards are openly available: the Kantesti AI engine blood-test benchmark on GitHub (https://github.com/emirhanai/kantesti-blood-test-benchmark) and the record on Figshare (DOI: 10.6084/m9.figshare.32095435). It is a self-run, automated benchmark, not independently validated or peer-reviewed.
Who reviews the medical content on Kantesti?
All medical content is written and reviewed by Thomas Klein, MD, our Chief Medical Officer. Dr. Klein is a board-certified clinical hematologist serving as Chief Medical Officer at Kantesti and AI-supported interpretation. Additional oversight is provided by our Medical Advisory Board.
Experience AI-Assisted Blood Test Analysis
Join millions of users worldwide who trust Kantesti's AI Blood Test Analyzer for AI-assisted, informational blood test interpretation in 75+ languages.
Corporate Transparency
We believe in full transparency about who we are and how we operate. Below you'll find our company registration details and leadership information.
As of March 2026, the Kantesti AI platform — currently operating as Kantesti V11 — is operated by Kantesti Ltd, a private limited company incorporated in England and Wales (UK Companies House No. 17090423), with its registered headquarters in London, United Kingdom. This UK incorporation consolidates global operations under a single, transparent corporate structure governed by United Kingdom regulatory standards, building on development work that began in 2019. The same proprietary neural network — interpreting more than 15,000 biomarkers across 75+ languages for over 2 million users in 127 countries — continues to be developed and maintained by the same engineering and medical teams. Our internal benchmarking, Medical Advisory Board oversight, and partnerships with Microsoft for Startups (Founders Hub), NVIDIA Inception, and Google Cloud remain unchanged. Users worldwide retain access to identical service quality, now backed by enhanced UK corporate governance alongside our existing GDPR (EU) and HIPAA (US) compliance.
Kantesti AI · Kantesti Ltd
Brand Names: Kantesti, Kantesti AI
Legal Entity: Kantesti Ltd (Private Limited Company)
Companies House No.: 17090423
Jurisdiction: England and Wales, United Kingdom
Business Type: AI Healthcare Technology (SaaS)
Founded: 2019 · UK Entity Registered: March 2026
Headquarters: 4 Raven Road, Unit 1c3-1100, London, E18 1HB, United Kingdom
Application Version: Kantesti V11
Contact: [email protected]
UK Phone: +44 7508 364740 (Mon–Fri 9:00–18:00 GMT)
DE Phone: +49 177 497 4039 (Mon–Fri 9:00–18:00 CET)
Leadership
Founder & CEO: Julian Emirhan Bulut
Visionary entrepreneur leading AI innovation in healthcare technology. Building accessible blood test interpretation tools for global health improvement.
Connect on LinkedInChief Medical Officer: Thomas Klein, MD
Board-certified clinical hematologist leading clinical oversight of our medical content.
Compliance
Data Protection: GDPR-aligned (EU)
Healthcare Privacy: HIPAA-aligned practices (US)
Medical Device: Not classified as medical device - Informational tool only
References & Standards
Our validation methodology and clinical standards are based on established medical guidelines and international standards.
- [1] World Health Organization (WHO). Use of Glycated Haemoglobin (HbA1c) in Diagnosis of Diabetes Mellitus. Geneva: WHO; 2011. Available from: https://www.who.int/diabetes/publications/diagnosis_diabetes2011/en/
- [2] International Organization for Standardization. ISO 15189:2022 Medical laboratories — Requirements for quality and competence. Geneva: ISO; 2022. https://www.iso.org/standard/76677.html
- [3] Clinical and Laboratory Standards Institute (CLSI). EP09c: Measurement Procedure Comparison and Bias Estimation Using Patient Samples. 3rd ed. Wayne, PA: CLSI; 2018.
- [4] National Institutes of Health (NIH). Blood Tests Reference Ranges. Bethesda, MD: NIH; Updated 2024. Available from: MedlinePlus
- [5] American Association for Clinical Chemistry (AACC). Laboratory Test Reference Ranges. Washington, DC: AACC; 2024. https://www.aacc.org/
- [6] International Federation of Clinical Chemistry (IFCC). Reference Measurement Procedures. Milan: IFCC; 2023. https://www.ifcc.org/
- [7] Bulut J E. Kantesti AI Engine — Blood Test Interpretation Benchmark (automated, open-source harness). Kantesti Ltd; 2026. Code & data: GitHub · DOI: 10.6084/m9.figshare.32095435
- [8] U.S. Department of Health and Human Services. HIPAA Privacy Rule. 45 CFR Part 160 and Subparts A and E of Part 164. Washington, DC: HHS; 2013.
- [9] European Parliament and Council. General Data Protection Regulation (GDPR). Regulation (EU) 2016/679. Brussels: EU; 2016.