Print and Probability Research Associate - Dietrich College

Carnegie Mellon University•Pittsburgh, PA

1d•Onsite

About The Position

Carnegie Mellon University's Dietrich College is seeking a Research Associate for the Print & Probability project. This role involves developing AI methods to identify printers of anonymous early modern books (1450-1800). Building on previous successes in identifying clandestine printers of famous works, this Schmidt Sciences-funded phase will integrate large language models (LLMs) with computer vision to uncover hidden networks of controversial printing during periods of censorship. The project emphasizes practical problem-solving, interdisciplinary learning, and collaboration.

Requirements

Master's degree required.
1-3 years of research experience required.
Demonstrated expertise with large language models (fine-tuning, prompting, deployment).
Strong Python programming with deep learning frameworks (PyTorch, TensorFlow).
Experience with unstructured historical data (text extraction, entity resolution, knowledge graphs).
Excellent communication skills and commitment to interdisciplinary collaboration.
Evidence of scholarly productivity (publications, presentations, software).
Applicants for this position must be currently legally authorized to work for CMU in the United States. CMU will not sponsor or take over the sponsorship of an employment visa for this opportunity.
Carnegie Mellon is not a qualifying employer for the STEM OPT benefit: only the 12-month OPT may be used to work at Carnegie Mellon.

Nice To Haves

Knowledge of early modern European history (1450-1800) or book history.
Experience with historical bibliography or archival research.
Familiarity with computer vision for document analysis.
Multilingual reading ability (e.g., English, Latin, French, Spanish, Italian, Dutch).

Responsibilities

Develop LLM-driven knowledge graphs that construct probabilistic historical priors from bibliographic records, trial transcripts, censorship lists, and apprenticeship data.
Design agentic frameworks using In-Context Learning and Chain-of-Thought prompting for transparent historical inference.
Develop Historical Hypotheses in collaboration with expert humanists and book historians.
Integrate top-down LLM hypotheses with established bottom-up vision pipeline (existing: dhSegment/Eynollah line extraction, damage detection models, 280M+ character image database).
Assist in original research on clandestine printing networks using computational tools.
Contribute to publications in both AI and humanities venues (machine learning conferences and book history journals).
Contribute to open-source tools and datasets for the research community.
Other duties as assigned.