Who we are
We started CliniNote because every research team we knew was solving the same problem manually: reading notes, copying values, and hoping the reviewer before them hadn't missed anything. There had to be a better way.
Our mission
Clinical evidence should not depend on how carefully a research coordinator annotated a chart on a Friday afternoon. It should be reproducible, auditable, and traceable to source documents, the same way laboratory results are.
That is the standard we are working toward. Every extracted entity CliniNote returns carries a source citation. Every output row has a confidence score. Every job can be re-run with the same pipeline version to verify reproducibility. We are not there yet on every dimension, but that is where we are going.
We are a small team, bootstrapped, based in Warsaw. We take on fewer customers than we could, because we want to actually understand each team's workflow before they go into production.
The team
CEO and Co-Founder
Spent years in clinical informatics at a large academic medical center in Central Europe, building and maintaining EHR integration pipelines. Left to start CliniNote after the third consecutive research project stalled over unstructured note data with no scalable solution.
CTO and Co-Founder
Computational linguist with a background in biomedical NLP research at a European university hospital consortium. Led annotation and model training on de-identified clinical corpora before joining Robert to build CliniNote's extraction engine. Holds a doctorate in computational linguistics.
Head of Clinical Science
Pharmacoepidemiologist who spent his career evaluating real-world medication safety at a pharmaceutical outcomes research organization. Joined CliniNote to ensure the extraction outputs meet the evidentiary standards that regulatory submissions require. Keeps Robert and Aleksandra grounded in what actually matters clinically.
How we got here
Robert and Aleksandra started working on the extraction prototype after leaving their respective institutions. Initial focus: does a fine-tuned clinical NLP model actually outperform general-purpose models on EHR notes? (It did, significantly, on the internal benchmark corpus.)
Partnered with an outcomes research team at a mid-sized European CRO to test CliniNote on a retrospective cardiovascular cohort. Processed 1,200 discharge summaries. First real accuracy validation: 88% F1 on primary diagnosis, 92% recall on documented medications.
Marcus joined as Head of Clinical Science to build the evaluation framework and ensure outputs meet pharmacoepidemiology standards. The team grew to three. Work began on the negation and uncertainty module, which became the feature most valued by early partners.
Launched the REST API alongside the web platform with project-level isolation, source citations, and configurable data retention. The first three paying customers onboarded within the first month. BAA process and EU-West data residency confirmed compliant with GDPR processor requirements.
Released the custom phenotype module, allowing teams to define study-specific clinical entities beyond the standard seven entity classes. The first use case: a rare disease research team building a cohort definition that no standard coding system covered. The feature became the main reason teams upgraded from Starter to Research.
We take early access calls ourselves, not a sales team. Tell us about your extraction challenge and we will give you an honest assessment of how CliniNote would handle it.
