Sequence prediction & protein design
Dr Stephan Breimann
PhD in computational biology at TUM. First author of our method papers, among them Nature Communications (2025).
About
Protein science, software, and the bench
A young company from Munich, built at the Technical University of Munich.
Software platform & automation
Amin Sohrabi
Doctoral researcher in computational biology at TUM, MSc in computer science. Previously co-founded and ran a remote-diagnostics company.
Mentors and advisors
Our mentors and advisors contribute expertise in protein design, bioinformatics, language models, agentic AI and entrepreneurship. Their research is linked by subject. They advise as individuals; their institutions do not endorse InXBio.
Generative protein design
Prof. Thomas Schlichthärle
Scientific mentor. AI-guided protein design, from minibinders to proteins that sense and reprogram cell signalling. Experience from the David Baker laboratory.
Bioinformatics
Prof. Dmitrij Frishman
Bioinformatics algorithms and software: structural genomics, proteomics, protein interactions and function prediction.
Protein language models
Dr Michael Heinzinger
Protein language models, and fine-tuning them for prediction tasks.
Agentic AI
Dr Sebastian Lobentanzer
Agentic workflows and knowledge management for biomedical research.
Entrepreneurship
Wolfgang Westermeier
Entrepreneurial mentor. Co-founder of Customized Drinks and FarmInsect.
How we work with you
Each project has agreed deliverables, acceptance criteria and a record of the decisions behind the result. Here is what that commitment covers.
A defined development step
We promise a defined development step against criteria we agree in writing before the work starts. A suitable minibinder is the development goal, not a guaranteed outcome.
If a development round misses the criteria
You get one further design and test series after release. You carry the external production and measurement cost; we carry our own development time and the documentation of what changed.
Laboratory work through project partners
Design, selection and construct engineering are ours. Production and binding measurement run through partner laboratories that we select for your assay and coordinate. We are still contracting those partners, so we name them per project rather than claiming a standing panel. The advantage of the arrangement is that the assay is matched to your experiment.
The design methods are public
Most of the models we use are openly available and anyone can run them. What we add is the selection: which method for which target and which experiment, why those candidates and not the others, and a record of both. The selection is made with our own set of explainable tools, paired with our agentic AI system.
We build what is missing around them
The published tools do not join up by themselves. Benchmarks, explainability tools and small applications that glue one step to the next are ours to write, and we write new ones for a customer's problem where none exists. The aim is a design process that is quicker, easier to read and easier to trust.
The agent works under supervision
The agent proposes and keeps the record; a person decides wherever a step costs money or material. We are building our own biosafety guardrails, and pairing the agent with our explainability work, so a result can be checked rather than trusted.
No animal experiments
We run no animal experiments. Where a customer's own programme generates omics data, we can analyse it.
Tell us the target that is holding your experiment up
You get a straight answer about what is realistic.
Research use only. Scope, criteria and deliverables are agreed per project.
