Takeda's Quantitative Clinical Pharmacology (QCP) group is exploring how open-source large language models (LLMs) and retrieval-augmented generation (RAG) pipelines can streamline pharmacometrics (PMx) workflows and regulatory query support as part of Takeda’s digital and data science transformation.
Requirements
- Strong programming skills in Python (experience with LangChain, HuggingFace, or PyTorch preferred).
- Familiarity with LLMs, embeddings, RAG architectures, and prompt engineering.
- Prior exposure to pharmacometric modeling (NONMEM, Monolix, or Stan) is desirable.
Responsibilities
- Develop a prototype RAG pipeline leveraging open-source LLMs, embeddings, and vector databases.
- Ingest and index internal clinical pharmacology and regulatory documents to enable natural-language search and summarization.
- Build workflows for automated NONMEM/Monolix code generation from structured specifications, validated against existing models.
- Design a chatbot interface for natural-language queries over indexed regulatory/QCP documents with grounded citations.
- Conduct evaluation of retrieval accuracy, code validity, and regulatory response quality.
- Document the architecture, methods, and findings for internal review and handover.
- Present results and recommendations to QCP leadership and Takeda stakeholders.
Other
- This position will be Hybrid (1–2 days/week in office) out of the Boston, MA location.
- Must be pursuing a PhD degree in Pharmacometrics, Quantitative Clinical Pharmacology, Computational Biology, Data Science, Computer Science, or a related field.
- Strong written and verbal communication skills, with the ability to translate technical concepts to scientific stakeholders.
- Self-motivated, detail-oriented, and able to work independently in a fast-paced, collaborative environment.
- Must be authorized to work in the U.S. on a permanent basis without requiring sponsorship
