Senior Data Scientist, Agentic AI Systems

Axle is a bioscience and information technology company focused on translational research, biomedical informatics, and data science applications for research centers and healthcare organizations. The Senior Data Scientist, Agentic AI Systems will build and operate conversational AI systems for rare disease research, owning major components from design through deployment and collaborating with NIH staff, clinical geneticists, and rare disease information specialists.

Responsibilities:

• Build agentic AI systems for rare disease research workflows. This includes the conversation logic, the rules that decide when enough information has been gathered, and the confirmation steps that catch a misreading before it reaches anything downstream
• Model outputs in Pydantic and use structured output and tool calling, so that every field a model produces is typed, validated, and traceable back to its source
• Write, version, and regression test the prompts behind clinical and scientific reasoning tasks. Prompts and output schemas are treated as code here, with tests to match
• Build evaluation for tasks that have no single right answer. Golden sets, offline regression suites, and model-based graders all have a place, and the results should be good enough to decide what ships
• Keep multi-step LLM workflows responsive under load. This covers async design, concurrency limits, streaming partial results to the client, and timeout and failure handling that holds up in production
• Log what the system does and why. Request identifiers, latency, errors, and the reasoning behind each automated choice all need to be captured, so that staff can review an AI-assisted result instead of taking it on faith
• Work out what researchers, clinicians, and patient communities need, and turn it into data models and system behavior
• Write the work up. You will contribute to manuscripts, conference abstracts, and posters with NIH investigators, and you will be credited as an author on work you helped produce