Senior XAI role posted

A new Cambridge, Massachusetts posting for a Senior AI Research Scientist in Explainable Artificial Intelligence asks for original research on how models reach biological conclusions, not just policy language about “responsible AI.” (jobrxiv.org) The listing, posted April 11, 2026, is for Merck & Co., Inc. and sits inside the company’s Data, AI, and Genome Sciences group, which says artificial intelligence and machine learning are central to drug discovery and biomarker work. (careercircle.com) (merck.com) Explainable artificial intelligence is the part of machine learning that tries to show why a model produced an answer, the way a lab notebook shows how a result was reached instead of just giving the result. The National Institute of Standards and Technology says explainable systems should provide reasons, make those reasons understandable, reflect the actual process used, and operate only within designed conditions. (nvlpubs.nist.gov) The Merck role spells that out in technical terms: the scientist would extract biological insight from foundation models, build post-hoc and intrinsic explainability methods, and create benchmarks, evaluations, and datasets tied to specific biological questions. The job also places that work inside a cross-functional team of computational biologists, data scientists, software engineers, and machine learning researchers. (careercircle.com) Interpretability and explainability are often used as if they mean the same thing, but research literature separates them. A 2024 paper on their relationship describes a split between interpretable models built to be understood directly and explainability techniques used to analyze more complex “black box” systems after the fact. (arxiv.org) That distinction shows up in the wording of the job itself. “Intrinsic explainability” points to models designed for transparency from the start, while “post-hoc” methods try to inspect a trained model afterward through tools such as feature attribution, counterfactuals, or saliency maps. (careercircle.com) (arxiv.org) The setting is drug research, where Merck says its data science work spans target discovery, research, and development, and where its Genome and Biomarker Sciences unit uses human genetics and genomics to identify targets and biomarkers. In that environment, an explanation is not just a user-facing feature; it can be part of deciding whether a model’s pattern matches disease biology closely enough to guide experiments. (merck.com) (jobs.merck.com) The posting also asks for senior-level scientific judgment. The duties include identifying research questions, defining data requirements, developing methods, and advancing therapeutic strategy, which puts the role closer to principal-investigator-style model science than to compliance review. (careercircle.com) Academic debate around post-hoc explanation remains active. A December 2024 position paper defended post-hoc explainability as a valid way to produce scientific understanding when it is paired with empirical validation and clear limits on what the explanation can claim. (arxiv.org) So the signal in this hiring post is concrete: one large drugmaker is recruiting a senior scientist to make foundation models legible enough for biology, with methods, benchmarks, and datasets treated as core research work. (jobrxiv.org) (careercircle.com)