Machine Learning Software Engineer

Variational AI•Vancouver, BC

55d•Remote

About The Position

Small molecule drug discovery is one of the most exciting open problems in machine learning. Traditional approaches require over ten years and two billion dollars to develop a new pharmaceutical, and their reliance on trial-and-error calls out for better predictive and generative models. The existent datasets are large enough to benefit from sophisticated deep learning architectures, but small enough that ML models can be trained in a few days, facilitating rapid experimentation and innovation. Nevertheless, the current industry standard has progressed little beyond shallow ML techniques and simple graph neural networks, largely due to the difficulty of integrating world-class machine learning research with chemistry and pharmacology expertise. Variational AI is building a generative foundation model for molecular structure and properties from the ground up. For over six years, we have been advancing the state-of-the-art, and delivering projects to customers including Merck, Rakovina and ImmVue Therapeutics. Variational AI is searching for a machine-learning software engineer to join us in our quest to radically accelerate the development of new drugs through machine learning excellence. You will help improve our existing code base by improving memory and compute efficiency; develop new automation pipelines for preparing datasets and running experiments; and generally move research into production. In this process, you will have the opportunity to build your skills by collaborating with our team of accomplished ML scientists and chemists. Software engineering expertise is the primary requirement, along with knowledge of deep learning fundamentals; knowledge of chemistry and pharmacology is preferred but not required.

Requirements

Degree in CS, applied mathematics, statistics, physics, or related discipline
Mastery of Python and PyTorch, including a deep understanding of tensor object structure and its performance implications, and the ability to write new, efficient tensor operations
Experience in one or more of: Python extension modules, Shell scripting, Code profiling, Large datasets, including storage formats and efficient loading, Dataframe libraries, SQL, Slurm or other HPC environments, CUDA programming, Cloud platforms, Containers
Two or more years of experience developing robust code on larger projects, including code review, refactoring, unit testing, version control, etc.
Intellectual curiosity and drive to excel