About
I'm a veterinarian and an NLP researcher, and the work is mostly about clinical records: how to label what's inside them reliably, at scale, when there is very little labelled data to learn from.
I started out as a systems engineer at Microsoft, went to vet school at Washington State, co-founded VetPronto and ran technology at Love That Pet, neither of which worked out, and then did a PhD at the University of Melbourne between the NLP group and the veterinary epidemiologists who run VetCompass Australia. At VIN I'm Chief AI Officer, and I built the first version of AI Mode, the assistant vets use when they're stuck on a case; it cites the source of every answer, and an engineering team runs it now. Through Datatroph, which I've run since 2008 with a bench of people I've worked with before, I take on fractional Chief AI Officer roles and technical due diligence; nearly all of that work is AI now, a good deal of it on clinical and genetic data, and some of it outside medicine. I'm also president of the Association for Veterinary Informatics, which is volunteer work.
Why
Too many of us will die earlier than we should, of things medicine already knows how to treat, because the system that delivers that knowledge can't learn from its own records: they go unread, the evidence in them goes unused, and the people who try to change that from the inside are more often worn down by it than not.
I don't think you can fix human healthcare from inside it, so my bet was that you could prove the fix somewhere with the same medicine and fewer barriers, which is most of why I became a veterinarian; the rest is that animals are worth helping on their own. Dogs and cats get many of the diseases we get and their records are just as detailed, but in a vet clinic you can change how antibiotics get prescribed and then look at what happened. If the bigger bet never pays off, the work still helps some dogs and cats and the people who love them, which is worth doing on its own.
That's what I've been doing since, some of it building and some of it helping other people build. I don't expect to be the one who solves it; I would like it solved.
Email Google Scholar LinkedIn GitHub
Currently
- Chief AI Officer, Veterinary Information Network
- Datatroph, a consulting firm I run
- President, Association for Veterinary Informatics
- Venture Partner and Scientific Advisor, Vetted Capital
Education
- PhD, Natural Language Processing and Veterinary Epidemiology, University of Melbourne, 2022
- DVM, Washington State University, 2011
- BS, Washington State University, 2008
Awards
- Best Lightning Talk, Australasian Computer Science Week, 2020
- Best Poster, Australian Veterinary Association Conference and CIS Doctoral Colloquium, 2019
Teaching
- Text and Web Analytics, Melbourne Business School, Master of Business Analytics, 2019
- Foundations of Computing, University of Melbourne, 2019
Service
- AAVSB AI Advisory Committee, 2024 to present
- ACM BioNLP program committee, 2021 to present
- Chair, NLP Reading Group, University of Melbourne, 2018 to 2019
- Cairns Turtle Rehabilitation Centre, volunteer, 2017
Talks
Neural Networks & NLP for Antimicrobial Stewardship
Australian National Antimicrobial Resistance Forum
NLP and VetCompass for Antimicrobial Usage Patterns
Australasian Computer Science Week (ACSW)
ML for Passive Data Extraction in Veterinary Practices
NCAS World Antimicrobial Awareness Week Webinar
Using AI to Explore Drug Usage from Clinical Records
Talbot Veterinary Informatics Symposium
Domain Adaptation for Veterinary Clinical Notes
BioNLP Workshop @ ACL 2020
State of the Art Diagnostics in Veterinary Medicine
Australian National Antimicrobial Resistance Forum
NLP & VetCompass for Antimicrobial Patterns
AVA Conference & CIS Doctoral Colloquium
Contrasting n-gram Matching and ClinicalBERT
N2C2/OHNLP Workshop @ AMIA