Pre-training large models that generalize well across the chemical space

The intern is expected to develop a pipeline for training large machine learning models on drug-like molecules to learn a broad understanding of biochemistry. This pipeline involves developing a model that can learn on many biological and quantum mechanical tasks concurrently while evaluating and improving the structure of the model using low-cost approaches from recent literature. Then, he will implement transfer learning to evaluate the model’s performance in predicting the properties of unseen molecules for unseen tasks. This last part is very relevant for the partner organization since it is expected to improve their ability to predict molecular properties on real-world tasks and speed up the discovery of novel drugs.

Faculty Supervisor:

Ioannis Mitliagkas

Student:

Partner:

Valence Discovery Inc

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Université de Montréal

Program: