You would be the first scientist in Valvian’s Computational Biology & AI function, working directly with the Head of Computational Biology & AI.
We are building a computational platform that brings together AI models and proprietary experimental data to find the targets and drug candidates in oncology and age-related disorders that the existing knowledge does not make obvious.
The first major challenge sits upstream of any AI model. In our therapeutic area, the evidence is scattered across different public resources, primary literature and internal experimental data, reported inconsistently, and deeply dependent on experimental context that is rarely recorded. Whether a data substrate is structured, provenance-aware and honest about uncertainty determines whether anything built on top of it can be trusted. A model trained on unstructured data across the literature learns the shape of that literature and its biases rather than the underlying biology.
That structured data substrate is the core of the role. What should the data model look like? Which observations can legitimately be compared? How should conflicting evidence be represented rather than quietly reconciled? How are different entities dependent on each other? How do we unify existing data with our own? You would help answer those and other questions, then implement the pipelines and processes to populate, maintain and expand this resource as it becomes a reliable substrate for the implementation of predictive AI models.
Because the function is new, the work is unusually broad. You would have active participation in developing the full platform, i.e. the data model, the pipelines, the curation, the automations, and the conversations with the bench, rather than owning one narrow slice of an existing system.