Guilherme Bauer-Negrini

Deep learning on 3D medical images: Developed 3D convolutional networks with transfer learning to classify and stage Alzheimer’s disease from volumetric tau-PET, training separate models per radiotracer and transferring representations between them to compensate for limited labelled data. Used leakage-safe subject-level cross-validation throughout. Secured funding as Principal Investigator through an Alzheimer’s Association Research Fellowship (USD $200,000) for “A deep learning tool to generalize tau-PET to predict Alzheimer’s disease progression.”

Image registration, harmonization & quality control: Co-developed Uniτ (uni-tau.com), a voxel-wise transformation placing images from four different tau PET radiotracers onto a single quantitative scale, including reference-region definition, spatial normalisation to common templates, and systematic QC of registration and segmentation across a multi-centre head-to-head cohort. Performed imaging processing and interpretation work for the HEAD Study, the largest Alzheimer’s disease methodological imaging study to date (co-author in The Lancet).

Population-scale genetics and longitudinal clinical records: Computed polygenic risk scores for 200,000+ participants and built cohort-definition and phenotyping pipelines over longitudinal electronic health records (NIH All of Us), testing genetic liability for ADHD as a predictor of incident mild cognitive impairment and dementia. First and co-first author in JAMA Psychiatry and JAMA Network Open.

Genetic risk linked to imaging and fluid biomarkers: In Alzheimer’s disease cohorts, related genetic liability to molecular and imaging outcomes and modelled joint structure across tau-PET, amyloid-PET, MRI, fluid biomarkers, and longitudinal cognition (co-author in Alzheimer’s & Dementia; Molecular Psychiatry).

Collaboration & dissemination: Worked across international institutions in the US, Canada, Brazil, Spain, and the Netherlands; presented as first author or speaker at major Alzheimer’s disease conferences (Human Amyloid Imaging, Alzheimer’s Imaging Consortium, Alzheimer’s Association International Conference); mentored junior researchers and contributed to grant applications.

São Paulo, Brazil · Remote

Data scientist in an R&D project funded by UNICEF to map overdue vaccinations and issue alerts to health workers; built healthcare data interfaces using HL7 FHIR; developed the containerized developer environment with Docker, poetry, and Bitbucket Pipelines; contributed to AI-enabled public-health alerting. The pilot project in Brazil involved 3,876 children and resulted in a promising initial increase in vaccine coverage, from 53% to 56%, a 3% improvement in just two months.