Etai Jacob, PhD

About

I lead R&D organizations at the intersection of AI, structural biology, and drug discovery and development. Over 20+ years spanning HiTech, Biotech, and Pharma, I’ve established and scaled pioneering teams from 10 to 40+ people—driving transformative advancements aimed at reshaping patient outcomes, accelerating drug development timelines, and enhancing R&D productivity.

My experience sits at the intersection of software engineering, computational biology, AI, and drug development—from embedded C/C++ engineering on custom ARM chipsets in HiTech, to computational structural biology and AI-driven drug target discovery in Biotech, to leading enterprise-scale AI research organizations pioneering novel methods and platforms across the full drug development value chain in Pharma. I champion large-scale strategic initiatives in highly matrixed environments, building trusted partnerships with global stakeholders and external collaborators—combining deep technical expertise with people leadership and business acumen.

Currently, I lead Applied Data Science and AI in Oncology R&D at AstraZeneca, where my group pioneers AI methods and enterprise platforms across the drug R&D pipeline—from the Predictive Biomarker Modeling Framework and Clinical Transformer to the enterprise-scale Biomarker Navigator platform—while leading enterprise-level partnerships to develop cutting-edge AI and foundation models for oncology. Recent work published in Cancer Cell, Nature Communications, and Nature npj Precision Oncology.

Experience

AstraZeneca

Head of Applied Data Science and AI, Oncology R&D

Feb 2025 – Present · Greater Boston, MA

Lead a global organization of 20+ data scientists, computational biologists, and AI researchers developing novel AI methods and platforms from early discovery through clinical development
Leading enterprise-level partnerships to develop cutting-edge foundation models for oncology—driving scientific strategy, evaluation frameworks, and large-scale clinicogenomic data integration
Architect agentic AI frameworks integrating LLMs with domain-specific scientific workflows

Senior Director, Head of Data Science and AI, Early Oncology

Jul 2022 – Feb 2025

Scaled the function to 40+ contributors. Led development of novel AI methods and platforms: PBMF (contrastive learning), Clinical Transformer (pretrained multimodal framework), and the enterprise-scale Biomarker Navigator platform for biomarker discovery and clinical development
Published in Cancer Cell, Nature Communications, Nature npj Precision Oncology. Senior/corresponding author

Director, Head of Data Science and AI, Early Oncology

Jul 2021 – Jul 2022

Built the data science and AI function from the ground up. Defined the strategic roadmap for AI/ML across oncology pipelines, driving portfolio-level investment decisions from target selection through late-phase clinical development

Previously: Lead for Statistics and Machine Learning, Early Oncology (Apr 2020 – Jun 2021)

Dana-Farber Cancer Institute

Lead Computational Biologist

2016 – 2020 · Boston, MA · Broad Institute & Harvard Medical School

Developed computational methods and production pipelines for single-cell genomics, multi-omics integration, and tumor heterogeneity
Built ML systems (deep neural networks, semi-supervised learning) for genomic analysis and translational research across 5+ labs

Compugen Ltd.

Senior Scientist → Project Leader & Algorithm Developer

2007 – 2016 · Tel-Aviv, Israel · Nasdaq: CGEN

Built computational discovery engines integrating protein structure, ML, and multi-omics data into production software platforms
Led cell-penetrating peptide discovery end-to-end—team management, budget, wet-lab validation, patent filing
Secured Israel Innovation Authority grants; established Rimonim consortium for RNAi therapeutics

Weizmann Institute / Bar-Ilan University

Research & Teaching Fellow

Israel · concurrent with Compugen

Novel methods for protein 3D structure prediction and discovery of aggregation prevention mechanisms
Published in eLife, PNAS, Cell Reports

Optimedia & HomenetIP

Software Engineer

Israel

HomenetIP: Designed and developed embedded software for smart-home residential gateway products on a custom ARM CPU and chipset. Performed low-level ARM processor optimization for performance-critical systems. C/C++ development with networking/IP stack and MMI components
Optimedia: Developed image processing and optimization algorithms and UI software for marketing and advertising applications using C++/VC++ (Windows). Built production software for real-time image analysis

Research Themes

Protein Structure Prediction

Novel correlated mutation analysis combining codon-level and amino acid information to predict inter-residue contacts.

Developed a new approach using combined analysis of amino acid and codon multiple sequence alignments to improve protein contact prediction.

Read the paper in eLife →

Protein Folding & Aggregation

Multi-domain protein folding via lattice models and Monte Carlo simulations. Discovered mechanisms for aggregation prevention.

N-terminal domains in two-domain proteins fold faster than C-terminal counterparts—reducing aggregation risk during co-translational folding.

Cell Reports → · PNAS →

AI for Drug R&D

Foundation model evaluation, predictive biomarker discovery with contrastive learning, pretrained transformers for clinical studies, agentic AI.

Leading enterprise-level partnerships to build cutting-edge foundation models for oncology. Developed the PBMF (15% survival improvement in Phase 3 IO trial) and Clinical Transformer.

Cancer Cell → · Nature Comms →

Translational Medicine

Enterprise AI platforms for biomarker strategy, multimodal data fusion, immune checkpoint biomarkers, and clonal hematopoiesis prediction.

Built the enterprise-scale Biomarker Navigator. Multimodal fusion for survival prediction. ML models for ICB therapy response. AI-based clonal hematopoiesis prediction from cfDNA.

npj Precision Onc (Multimodal) → · npj Precision Onc (CHIP) → · Cancer Res (Biomarker Navigator) →

Selected Publications

Featured: Predictive Biomarker Modeling Framework

A contrastive-learning neural network framework that discovers predictive biomarkers from clinicogenomic data, improving survival outcomes by 15% in a Phase 3 IO trial. Integrated with the Clinical Transformer to span the full drug development lifecycle.

▶ Watch the explainer video · Cancer Cell paper

From Counterfactual Scenario Simulations of Drug MoA to Patient Stratification Using Foundation Models

A pretrained transformer framework for multimodal clinicogenomic data that outperforms state-of-the-art survival prediction methods. The model leverages self-supervised, gradual, and transfer learning to handle sparse clinical datasets, generates digital twins for in silico perturbation experiments, and extracts interpretable functional modules that reveal biological mechanisms of drug response.

▶ Watch the explainer video · Nature Communications paper

AI/ML for Drug Discovery & Translational Medicine

AI-driven predictive biomarker discovery with contrastive learning. Cancer Cell2025 doi

Pretrained transformers improve predictions of treatment efficacy. Nature Communications2025 doi

AI-based model for prediction of clonal hematopoiesis variants in cfDNA. Nature npj Precision Oncology2025 doi

Multimodal data fusion for survival prediction. Nature npj Precision Oncology2025 doi

ML modeling of patient health signals on ICB therapy. iScience2024 doi

Protein Structure & Computational Biology

Codon-level information improves protein contact predictions. E. Jacob, R. Unger, A. Horovitz. eLife2015 doi

Cross-linking reveals laminin coiled-coil architecture. G. Armony, E. Jacob et al. PNAS2016 doi

N-terminal domains fold faster than C-terminal counterparts. E. Jacob et al. Cell Reports2013 doi

Additional publications in Cell (2018), Cell Signalling (2019), Bioinformatics (2007). Full list on Google Scholar.