Gregory Berlinerblau

Lead Machine Learning Research Scientist & Head of Computation (Department of Astrophysics) at The Hebrew University - Racah Institute of Physics

Designed Transformers, Temporal Convolutional Networks (TCNs), and Graph Neural Networks (GNNs) for time-series prediction of supernova events on multi-terabyte datasets. Leveraged large language models (LLMs) on AWS with literature-based feature augmentation and retrieval-augmented generation (RAG) to enhance low-data astrophysical predictions. Accelerated model training by 220% by building reproducible HPC/Docker pipelines and optimizing workflows. Provided computational support across seven astrophysics research groups; mentored PhD researchers in ML, HPC, and workflow optimization; developed Python pipelines and accelerated experiments and data analysis.

Machine Learning Research Scientist (Drug Discovery) at Drug Discovery AI Research Project

Invented and implemented a novel, large-scale ML pipeline for protein- and affinity-specific ligand generation (0.5M+ compounds), combining Transformers, embeddings, contrastive learning, and multi-omics data. Explored literature-guided molecular generation using semantic embeddings and scientific text mining (RAG) to enrich manifold conditioning. Designed end-to-end reproducible workflows for PDBbind and ChEMBL datasets, including ADMET and property prediction, enabling high-throughput generation of high-affinity candidates. Led cross-disciplinary work integrating AI with domain experts and disseminated findings via open-source tools and blogs.

Machine Learning Researcher (Cybersecurity) at Acronis

Reduced false negatives by researching, benchmarking, and deploying state-of-the-art ML models (GNNs, autoencoders, Siamese networks, KNN) on large-scale datasets, improving overall system accuracy. Designed and implemented reproducible, scalable ML pipelines, integrating data preprocessing, model prototyping, and validation workflows to ensure production-grade performance. Applied advanced ML theory to optimize architectures and training strategies, evaluating trade-offs across diverse tasks. Collaborated across global offices to investigate ML workflows and data structures, informing future modeling approaches and cross-disciplinary projects. Directed end-to-end model validation and testing, including benchmarking and hyperparameter tuning.

Head of Machine Learning Research (Principal Research Scientist, Biomedical AI; cross-departmental lab lead) at Sheba Medical Center - Diagnostic Neuroimaging Lab

Achieved 85% accuracy in classifying TBI, depression, and other brain abnormalities by developing novel CNNs on MRI/fMRI connectome data; led to a patent. Invented a real-time cerebral blood clot detection system (YOLO + VAE) in collaboration with neurosurgeons, supervising labeling, enabling transfer learning with angiography datasets, and prototyping for clinical research and startup applications. Streamlined research infrastructure with high-performance servers and automated MRI/fMRI pipelines; provided hands-on training and mentored PhD students; led interdisciplinary teams and contributed to patent filings and peer-reviewed publications.