OccamzRazor

The data and modeling team behind research in neurodegeneration.

OccamzRazor builds the infrastructure that transforms data into impactful research.

Contact us → What we do →

More data than the field can use.

Neurodegeneration research now produces imaging, multi-omics, digital sensor streams, and longitudinal clinical data at a scale no single lab was built to handle. Most of it is collected once, analyzed once, and never joined to anything else. Cohorts can’t talk to each other. Wearable data sits in vendor formats. Models are trained on one site and never tested on another.

The bottleneck is no longer instruments or ambition. It is the unglamorous layer between collecting data and learning from it: harmonization, interoperability, validated models, and people who can run them.

That layer is what we build.

Infrastructure, models, and the people to run them.

1/

Data infrastructure for research consortia

We ingest, harmonize, and document heterogeneous datasets across sites, devices, and diseases so they can be reused, shared, and modeled. Standards-first, built to outlast the grant that funded them.

Typical workCross-site data models, device-agnostic wearable pipelines, provenance and versioning, interoperability with the data standards your consortium already uses, and documentation a new analyst can pick up in a day.
2/

Biomarker discovery and predictive modeling

From raw sensor signal to candidate endpoint. From small, noisy datasets to models that survive validation. Parsimony is the house rule: the simplest model that holds up is the one we ship.

Typical workDigital biomarker development from wearables, progression modeling, patient stratification, predictive models for costly experimental decisions, and held-out and external validation.
3/

An embedded data-science team

Most neurodegeneration groups don’t need another platform. They need people who can make the data they already have work. We embed with your scientists, on your data, under your governance, and leave behind code and documentation your team can run without us.

Selected work.

LUMC / ProPark consortium · Leiden

Wearables as biomarkers for Parkinson’s

The ProPark Data Platform → npj Parkinson’s Disease, 2025 →

ProPark is a multi-year observational study led by Leiden University Medical Center with sites across the Netherlands, following roughly 900 people with Parkinson’s alongside healthy controls — in the clinic and, through repeated week-long wearable recordings, at home. Since 2023 we have been the consortium’s data engineering partner: we designed, built, and operate the platform that receives its clinical, questionnaire, medication, and sensor data, processes it in a controlled compute environment, and serves it to researchers through consent-governed, queryable views. More than a trillion data points, every transformation logged, all storage and processing in the EU. Cohort-wide tremor feature extraction that took twelve weeks now runs in under thirty minutes, and the wearable pipelines underpin the study’s published work on assessing tremor at home.

What it showsWe can build the data layer for a multi-site neurodegeneration consortium and turn continuous sensor data into candidate endpoints.

Predictive modeling for cell line development

Outside neuroscience, the same discipline. For a top-20 pharmaceutical company, we built machine learning models that predict clone performance in cell line development, helping their scientists decide earlier which candidates to carry forward. Details are under confidentiality; we are happy to discuss the approach directly.

What it showsOur models meet the validation and delivery standards of a global pharmaceutical company.
The Parkinsome · Knowledge extraction

Mapping what science knows about Parkinson’s

NeurIPS 2020 · Relation-weighted link prediction → NeurIPS 2020 · Biomedical information extraction → ACL 2018 · Training classifiers with natural language explanations →

Before “AI co-scientist” was a phrase, we built one for Parkinson’s. Our knowledge-extraction platform read the literature at scale and integrated it with structured datasets into a unified map of the disease — work published at NeurIPS 2020 and ACL 2018 and supported by the Michael J. Fox Foundation.

What it showsA decade of neurodegeneration focus, and methods that are still relevant to the consortium tools now being built.

We started in Parkinson’s a decade ago and never left.

The diseases of brain aging affect more than fifty million people, have no cures, and are studied one at a time in silos that don’t share data. The technologies that could change that — sensors, omics, imaging, machine learning — are already generating the evidence. What’s missing is the engineering and the people to make it usable across diseases.

We think the next decade of progress in neurodegeneration will be decided less by any single discovery than by whether the field can learn from the data it already has. That is the problem we chose.

Who you’d be working with.

Robbie Narang, PhD

CEO

Robbie leads OccamzRazor’s partnerships and delivery. He was Chief Operating Officer at Notable Labs, served as a Congressional Innovation Fellow on the House Ways and Means Subcommittee on Health, and spent over a decade in diagnostics development and laboratory automation for drug discovery. He holds a PhD in Molecular Microbiology from Tufts University.

Katharina Sophia Volz, PhD

Founder & Executive Chair

Katharina founded OccamzRazor in 2016, led it as CEO, and now chairs its board. She built the company around the Parkinsome, the knowledge-extraction work that mapped the published science of Parkinson’s. Before OccamzRazor she was a stem cell biologist, with research at Harvard Medical School, UCLA, the Howard Hughes Medical Institute, and Stanford, where she was the university’s first PhD in Stem Cell Biology and Regenerative Medicine. She was named to Forbes 30 Under 30 in Science and MIT Technology Review’s 35 Innovators Under 35.

Partners

Describe the dataset and the question.

We work with research consortia, foundations, and pharma teams that have neurodegeneration data and need it to do more. Write to us and describe the dataset and the question.