Intern - Translational Safety - Computational Toxicology 🎓

Development Sciences (DevSci) spans the entire drug discovery and development cycle — from early stage research to drug commercialization. Part of the drug development pipeline in DevSci includes the preclinical safety evaluation of candidate therapeutic molecules by toxicologists and pathologists in the Translational Safety (TS) department in order to enable further evaluation in humans. Translational Safety is an integral part of DevSci. We contribute to the organization’s success by providing scientific insights and ensuring the safety of molecules that advance through the pipeline to patients. We do this to support the DevSci vision to deliver the right drug in the right dose to the right patient. We are also committed to providing better outcomes for our people, patients, business, and communities by advancing and boldly championing diversity, equity, and inclusion in our work.

The Translational Safety organization is composed of several integrated sub-functions. This summer intern project falls within the Computational Toxicology sub-function.The Computational Toxicology group enables early and accurate compound safety profiling by leveraging all relevant data (in vitro, ex vivo, in vivo), advanced analytics and computational modeling while closely working with other Translational Safety subfunctions such as Investigative Toxicology and Complex In Vitro Systems and a few subfunctions within gRED Computational Sciences Center of Excellence (CS CoE) organization.

This internship position is located in South San Francisco, on-site.

The Computational Toxicology group is seeking a talented summer intern to expand the capabilities of internally developed LLM-powered AI agents by building a critical new module that enables natural-language access to gene expression and cell-type context resources, leveraging public and internally curated datasets.

The intern will build a robust agentic workflow that supports users by providing fast, consistent biological context for targets and pathways:

• Enable user-friendly queries that return quantitative tissue- and cell-type expression summaries, with robust filtering over key metadata (e.g., tissue, cell type, annotation, condition).

• Develop MCP-based tools and prompt/context patterns that connect the agent to expression datasets and execute the underlying retrieval reliably.

• Perform practical data harmonization to support consistent cross-dataset interpretation (e.g., feature/metadata standardization; gene set/pathway queries).

• Integrate the module into the existing agent experience and standardize outputs for scientific review, including basic visualization (tables/plots).

• Develop a lightweight evaluation/testing framework based on representative stakeholder questions, and produce basic training materials and usage examples.

The overarching goal is to democratize access to expression and cell-type context evidence for toxicologists, pathologists, and predictive toxicology scientists, accelerating target-related interpretation and follow-up.