NicheFlow learns an underlying numerical representation of species' niches, letting you do neat things like interpolating between two species, generating hypothetical niches of non-existent species. This animation also shows how the model treats environmental and geographic space separately.

I am currently training NicheFlow 2.0 on several hundreds of thousands of species' range maps and will be releasing it publicly in the coming months.

Join me at the #ESA2025 meeting in August to learn more about it. Stay tuned for an announcement of my #ESA2025 short course on foundation models.

The magic? NicheFlow uses generative AI to:
🌍 Share knowledge across species
🧬 Capture complex environmental relationships
📊 Work with limited data
🔮 Predict distributions for rare species

Preprint: biorxiv.org/content/10.1101/2024.10.15.618541v1

P.S. 'It's not magic.'

For more details on graduate opportunities in my lab see: rdinnager.github.io/dinnage_lab_...

I am also actively recruiting graduate student for my new Assistant Professor role starting in July at the University of Alberta.

If you will be at #Evol2025 and are looking for opportunities to study at the intersection of AI/ML and biology (especially evolution), let's talk in Athens!

Currently funded by NSF to develop AI that learns ecological principles from species distributions. Looking forward to connecting with the computational biology community here!

Lab: rdinnager.github.io/dinnage_lab_website/
Code: github.com/rdinnager

3/3

Recent highlights:
📚 NicheFlow: First foundation model for species distributions
🌸 PhenoVision: AI analyzing 53M field photos (98.5% accuracy!)
📦 Open-source R packages: slimr, ENMTools, phyr

I believe in democratizing computational methods for all biologists.

2/3