Scholar
Eugene Brevdo
Google Scholar ID: NvMCACEAAAAJ
Google Inc.
Citations & Impact
Citations
32,279
H-index
25
i10-index
28
Publications
20
Co-authors
17
list available
Publications
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
- Involved in multiple projects including MLGO, a framework for integrating ML techniques systematically in LLVM; ProtNLM, a new method used by UniProt to automatically annotate uncharacterized protein sequences; Reverb, an efficient and easy-to-use data storage and transport system designed for machine learning research. Also a contributor to TensorFlow, TensorFlow Probability, TF Agents, and other open-source projects.
Research Experience
- Staff Software Engineer at Google DeepMind, focusing on protein understanding and optimization under uncertainty. Multi-stage peptide library design, co-optimizing cell permeability and protein binding. Developed LLM models for protein function annotation and target-conditional optimization. Co-TLM of the TF-Agents team, TLM of the Brain Learned Systems Team, built smarter query optimizers, cache eviction algorithms, etc.
Education
- PhD in Electrical Engineering, 2011, Princeton University, advisers Peter Ramadge and Ingrid Daubechies; BSc in Electrical, Computer, and Systems Engineering, 2005, Rensselaer Polytechnic Institute.
Background
- Research interests span many interconnected areas: Optimization under uncertainty/constraints and experimental design for e.g., software systems and high throughput screening in biology. Software systems for training and deploying ML, Bandits, and RL models. Machine Learning applied to optimizing large software systems (databases, datacenter scheduling, caches, compilers like LLVM and XLA/TPU).
Miscellany
- Worked at Lifecode, Inc., building supervised learning ML pipelines for clinical diagnosis of rare diseases from NGS assays.
