Consulting
Advice where machine learning meets the clinic.
Alongside clinical and research work, I am available for a small number of selected engagements — for teams that need single-cell and machine-learning depth together with a physician’s sense of what matters for patients.
- Availability
- Open to selected engagements, alongside clinical and doctoral work
- Formats
- Advisory calls · Project-based work · Ongoing advisory
- Setting
- Remote, or in person in Munich
- Languages
- English · German · Portuguese
Service areas
- 01
AI in biomedicine
Choosing, building and stress-testing models against real biological and clinical data.
- Model and architecture review — graph neural networks, transformer-based models and simpler baselines, matched to your data and question
- Benchmark design: baselines, metrics and data splits that make comparisons fair and results reproducible
- Training and evaluation setups in PyTorch, PyTorch Lightning and PyTorch Geometric, with experiment tracking in Weights & Biases
- 02
Single-cell & spatial omics strategy
An analysis plan that starts from the biological question — before the first sample is sequenced, or once the data is in.
- Study-design input for single-cell experiments: samples, controls and the metadata you will need later
- Analysis-plan and code review for single-cell and spatial data across scanpy, squidpy and scvi-tools
- Data-integration and reference-mapping strategy, including feature selection for atlas-scale data
- 03
Computational biology
Reproducible pipelines for sequencing data — documented, versioned and built to be handed over.
- Workflow design and review in WDL, combining deep-learning models with established bioinformatics tools
- Whole-exome and transcriptomic analyses, including mitochondrial and nuclear variant calling for chimerism and mosaicism
- Immunogenomics questions, including minor histocompatibility antigen prediction in transplantation cohorts
- 04
Clinical translation
Keeping computational work anchored in clinical reality — patients, endpoints and how care is actually delivered.
- Clinical input on computational projects in hematology and oncology: cohorts, endpoints and confounders
- Evaluation of AI and large-language-model tools for clinical decision support and patient communication
- Translating between data-science and clinical teams, so that questions and results are framed for both
Who it’s for
Teams working where biology, data and patient care meet.
Biotech & pharma R&D
Translational and computational teams working with single-cell, spatial or clinical data.
Academic labs & consortia
Method choice, benchmarking and analysis strategy for multi-sample and atlas-scale projects.
Clinical & translational groups
Bringing machine learning to patient cohorts with a clear clinical question.
AI & health-tech start-ups
Scientific and clinical feedback on models, data and evaluation setups.
How an engagement works
Step 01
Intro call
A short conversation about your question, data and constraints — and an honest view of whether I am the right fit.
Step 02
Scoping & proposal
A written scope with goals, deliverables and timeline, and what is explicitly out of scope.
Step 03
Project or advisory
A focused, time-boxed project, or light-touch ongoing advice with regular check-ins.
Step 04
Handover
Documented code, decisions and recommendations that your team can own and build on.
A question that needs both the data and the clinic?
Tell me briefly about the question, the data and the timeline. I will reply with an honest view of whether — and how — I can help.