A RARE CLINICAL + COMPUTATIONAL TRAJECTORY
Jacob Shreve is best understood not as a clinician who later became curious about artificial intelligence, and not as a computer scientist observing medicine from a distance, but as a physician–computer scientist whose career has genuinely spanned both domains.
His early professional work was rooted in bioinformatics, genome analysis, software-intensive research, and computational biology. He then moved into medicine to bring that technical background closer to patient care, ultimately training in internal medicine at Cleveland Clinic and in hematology/oncology at Mayo Clinic.
Today his work centers on clinically grounded AI in oncology: PET/CT radiomics, computer vision, multimodal predictive modeling, and translational precision medicine in hematologic malignancies and related cancer problems. Across writing, speaking, and software work, the through-line is consistent: models should be rigorous, externally credible, and useful in real clinical workflows.
Bioinformatics, genomics, and software development were central long before the current wave of clinical AI.
Medical school, Cleveland Clinic residency, and Mayo Clinic fellowship grounded that technical background in real oncology practice.
The resulting perspective is unusually practical: build models that are rigorous, reproducible, and deployable in real cancer workflows.
OVERVIEW
For collaborators and technical reviewers, the key point is that the computational identity here predates the current AI wave. That makes the oncology-AI work feel earned, durable, and clinically literate rather than opportunistic.
Before medical training, Jacob worked in computational biology and software-driven research across genome analysis, transcriptomics, viral genomics, and pipeline design. That early work established fluency with data architecture, reproducible analysis, and scientific computing long before “AI in medicine” became fashionable.
Training at Cleveland Clinic and Mayo Clinic added the practical realities that many AI projects miss: staging, treatment-response assessment, toxicity, workflow friction, and the high evidentiary bar clinicians expect before adopting a model.
At Mayo Clinic, he founded the Hematology AI Group and helped push early AI efforts toward usable hematology and oncology workflows. That physician-led mindset now carries directly into Ingensia’s product and research design.
RESEARCH FOCUS
The strongest public through-lines in Jacob Shreve’s work are consistent across publications, talks, and software development: clinically grounded AI in oncology, PET/CT radiomics, multimodal prediction, and a preference for methods that can survive real clinical scrutiny.
Automated lesion segmentation, radiomics extraction, and image-derived biomarker work in lymphoma and multiple myeloma, with quantitative imaging features used for risk stratification and survival-associated modeling.
Rather than treating imaging as a silo, the broader program connects scan-derived features with laboratory, clinical, and—in selected settings—molecular variables to build clinically meaningful predictive models.
Diffuse large B-cell lymphoma, multiple myeloma, myelodysplastic syndromes, acute myeloid leukemia, and cellular therapy outcomes form the most coherent disease-oriented body of work.
Repeated public emphasis on external validation, reproducibility, and workflow fit distinguishes this work from generic AI marketing. The goal is not merely to produce impressive metrics, but models that remain credible in real use.
Clinical adoption depends on external credibility. Models that perform well only in narrow retrospective settings are not enough.
The strongest model is still limited if it cannot be packaged into a realistic pipeline, reviewed by clinicians, and integrated into day-to-day practice.
Cancer care rarely depends on one data stream. Imaging, clinical variables, and molecular context often become more informative when modeled together.
SELECTED PUBLICATIONS
This is a curated set chosen to show range: foundational AI-in-oncology thinking, clinically serious predictive modeling, PET/CT radiomics in hematologic malignancies, and the earlier computational research base that preceded medical training.
A flagship review article that frames artificial intelligence in oncology with clinical realism, technical breadth, and attention to implementation barriers.
View publicationLarge multi-institutional machine-learning work in myelodysplastic syndromes, useful as evidence of clinically serious prediction modeling rather than conceptual AI commentary alone.
View publicationAn integrated diagnostic modeling paper that ties genomic and clinical information together—an important proof point for multimodal translational work.
View publicationA direct anchor for the imaging-AI identity: automated PET/CT segmentation paired with downstream radiomics for DLBCL risk stratification.
View publicationShows the imaging program extending beyond baseline description into delta-radiomics and treatment-response trajectory analysis.
View publicationExtends the PET/CT radiomics work into multiple myeloma and supports the broader point that the imaging program is not limited to one disease setting.
View publicationIllustrates movement from prognosis into treatment-response prediction and further reinforces the disease-focused hematology AI thread.
View publicationAn early computational publication that helps explain why the current AI work has unusually deep software and pipeline roots.
View publicationTHOUGHT LEADERSHIP
For outside collaborators, invited speaking and public writing are often useful trust signals because they show how someone thinks in front of a sophisticated audience—not only what appears on a CV.
Keynote-level speaking on artificial intelligence in oncology, with emphasis on current capabilities, future opportunities, and the standards needed for responsible adoption.
Invited presentation work around AI in oncology and the realistic clinical utility of emerging predictive models.
Publicly surfaced “AI in Oncology: Prime Time” presentation, showing a sustained public-facing role in explaining the field to practicing clinicians.
Founder of an AI Interest Group and the Mayo Hematology AI Group, with ongoing peer-review and scholarly service across AI-related manuscripts and grant activity.
SHORT BIOS
Included here because they are often useful for conference materials, collaboration summaries, or concise introductions.
Jacob T. Shreve, MD, MS, is a hematologist/medical oncologist and physician–computer scientist whose work sits at the intersection of cancer care, bioinformatics, and artificial intelligence. Trained at Purdue University, Indiana University School of Medicine, Cleveland Clinic, and Mayo Clinic, he has built a career focused on clinically grounded AI in oncology, including radiomics, multimodal predictive modeling, and precision medicine. His work emphasizes reproducibility, external validation, and practical integration into real clinical workflows.
Jacob T. Shreve, MD, MS, is a hematologist/medical oncologist and physician-scientist with an unusually deep foundation in bioinformatics, software-driven research, and translational artificial intelligence. Before entering medicine, he worked across academic and private-sector computational biology, contributing to genome analysis, pipeline development, and viral genomics. He later completed medical training at Indiana University School of Medicine, internal medicine residency at Cleveland Clinic, and hematology/oncology fellowship at Mayo Clinic.
His research portfolio spans precision medicine, machine learning in hematologic malignancies, PET/CT radiomics, computer vision, and multimodal clinical prediction. He has authored and coauthored work ranging from genomics and virology to myelodysplastic syndromes, acute myeloid leukemia, oncology outcomes prediction, and AI in cancer care. Across scholarship and public talks, he is known for advocating AI that is rigorous, reproducible, externally validated, and genuinely usable in clinical practice.
LET’S BUILD A STUDY OR DEPLOYMENT PLAN
Use this site as the front door for retrospective cohort studies, biomarker discovery, external benchmarking, and custom PET/CT analytics workflows.
