AI Research Scientist

Dayhoff Labs•Cambridge, MA

14h

About The Position

Design and train frontier models across biology and chemistry, including molecular structure and dynamics, reactions, and whole reaction networks. Model predictions will inform wet-lab experiments, while experimental findings will guide subsequent model iterations. The role involves training large models for enzyme–substrate prediction, neural network potentials, and inverse design of reaction networks, with end-to-end ownership of model architecture, data pipelines, training, debugging, and evaluation. The scientist will work directly with chemists and biochemists, communicate across disciplines, and incorporate new data into each iteration. The team is small, diverse, and global, and works quickly through shipping and iteration. The company is focused on reverse-engineering the origin of life and applying insights into molecular networks to areas such as catalysis, green synthesis, synthetic biology, and programmable matter.

Requirements

Demonstrated experience training large models end to end, with the ability to discuss in detail what failed and how it was fixed.
Strong machine-learning engineering fundamentals, including architectures, training dynamics, data pipelines, and evaluation.
Prior experience working directly on a chemistry, biology, or related physical-science problem.
Ability to clearly communicate complex technical concepts to colleagues whose primary background is biology or chemistry rather than AI.

Nice To Haves

PhD in a quantitative field; the posting notes that depth matters more than credentials.
Experience in ML for chemistry or biology, such as neural network potentials, graph neural networks, protein models, or reaction models.

Responsibilities

Train large models for enzyme–substrate prediction, neural network potentials, and inverse design of reaction networks.
Own models end to end, including architecture, data pipelines, training, debugging, and evaluation.
Work directly with chemists and biochemists and communicate across the fields of AI, chemistry, and biology.
Fold new data into each model iteration.