I'm an experienced data & machine learning scientist, with a focus on applied machine learning (ML) and natural language processing (NLP). Currently, I'm a Principal Applied Scientist at Xero, where I lead the ML Products team in Toronto, Canada.
I was previously a Senior Data Scientist at Citrine Informatics, where I led customer-facing and internal R&D teams applying ML to the materials and chemicals industries. My PhD in Materials Science & Engineering was awarded from MIT, and my thesis pioneered the use of several NLP techniques towards advanced materials discovery.
I enjoy cooking 🍜, martial arts 🥋, and playing video games 🎮. I also love dogs 🐕 and exploring the outdoors ⛰️.
I am a co-inventor of this US Patent, which introduces new metrics for evaluating the potential for success of materials discovery R&D projects.
In this patent, a system is described for leveraging machine learned predictions towards inverse design of materials. If choosing between some number of potential R&D scenarios, metrics computed from machine learned predictions can be shown to accurately determine which scenarios are more likely to lead to the discovery of superior materials.
As the lead author on this peer-reviewed journal article, I show how deep generative models leverage historical data and inform synthesis methods for novel materials.
Conditional Variational Autoencoders are combined with NLP techniques (word embeddings from language models & sequence labelling models for named entity recognition) to learn a generalized model for inorganic materials syntheses. Querying and inspection of the model reveals the successful capture of existing domain knowledge, and the potential to generate synthesis recommendations for novel materials.
See the code & data on GitHub
As the lead author on this article in the inaugural issue of the journal Matter (by Cell Press) I outline the current challenges and opportunities in reporting and writing standards for materials synthesis methods in the academic community.
The relationship between human-centric writing and readiness for AI-driven methodology is highlighted as a motivating example.