Machine Learning for Computational Materials Science
Description
As a complement to our physics-based simulation models, we are developing machine learning models for predicting properties of materials and formulations as a function of composition and physical conditions. To build an efficient lightweight property prediction model for individual prediction tasks we rely on off-the-shelf foundation models for chemistry and train a small prediction head. We have demonstrated a proof of concept of this solution on liquid electrolyte conductivity but see a potential for broad generalization to other industrially relevant use cases.
To build a model like our PoC for additional use cases there is a need to 1) identify relevant available data sets and academic research papers, 2) extract/ingest, clean and standardize the data, 3) train the model, and 4) deploy the model.
We are looking for a machine learning and MLOps research intern (research-focused internship), who will be involved in benchmarking, improving and relieving bottle necks in this pipeline. We expect responsibilities to straddle data extraction, model development, MLOps and process optimization.
Responsibilities:
- Establish baseline times required to do the full pipeline from identifying relevant datasets and research papers to deploying an improved model.
- Identify and alleviate the critical bottlenecks in the pipeline.
- Automate steps to improve the efficiency of the full pipeline
- Explore alternative foundation models and hardware choices
- Optimize inference for speed and cost efficiency.
Skills
Sign in to read and join the discussion.