ML Scientist: Synthesis Design for Drug Discovery

Genentech; San Francisco (CA); On-site; USD 148,000 - 274,000; Full time

Job summary

Genentech, part of Roche, seeks a Machine Learning Scientist to advance synthesis-aware molecular design in the AI for Drug Discovery team in San Francisco. You will develop ML methods spanning retrosynthesis, synthesis planning, and molecular generation to close the loop with automated synthesis.

You should have a PhD in a quantitative field, hands-on PyTorch/JAX, and experience with graph neural nets, reinforcement learning, and cheminformatics tools.

Qualifications

• PhD or equivalent in ML, computational chemistry, chemical engineering, or related field.
• Strong publication record or portfolio of projects.

Responsibilities

• Develop ML methods for synthesis-aware molecular design across retrosynthesis, synthesis planning, molecular generation, and search in synthesizable chemical spaces.
• Integrate proprietary reaction and biochemical data to design the next generation of synthesis-aware models and workflows for hit finding and optimisation.
• Build robust, scalable pipelines for active-learning loops that interface directly with automated and high-throughput synthesis platforms.
• Design novel batch synthesis-planning algorithms that maximise chemical-space coverage, information gain and experimental efficiency.
• Drive scientific impact through publications, open-source releases, and conference talks.
• Collaborate widely with computational and experimental researchers at Roche and with academic partners.

Skills

Graph Neural Networks

Sequence Models

Reinforcement Learning

Python

PyTorch

JAX

RDKit

Openeye

PhD

Publications

Education

PhD in ML/Computational Chemistry

Tools

RDKit

Openeye

Git/GitHub

Job description

Genentech, part of Roche, seeks a Machine Learning Scientist to advance synthesis-aware molecular design in the AI for Drug Discovery team in San Francisco. You will develop ML methods spanning retrosynthesis, synthesis planning, and molecular generation to close the loop with automated synthesis.

You should have a PhD in a quantitative field, hands-on PyTorch/JAX, and experience with graph neural nets, reinforcement learning, and cheminformatics tools.