In this live webinar on Thu, Mar 12, 2026 11:00 AM EST | 5:00 PM CET, we’ll introduce Model Atlas- a new way to chart and navigate large-scale model repositories as a connected graph of models, attributes, and fine-tuning relationships.
Yotam Azriel (CEO & Co-Founder, Tensorleap) will host Eliyahu Horwitz, creator of Model Atlas, for a practical, research-backed walkthrough- grounded in real examples from computer vision.
Public model hubs now contain millions of models, many lightly documented, making discovery, comparison, and ecosystem-level analysis far harder than it should be. Model Atlas turns that sprawl into an interactive map you can reason about, revealing how models cluster, evolve, and branch through fine-tuning and lineage.
In this webinar, we’ll cover:
Two computer vision applications: spotting emerging trends in CV model design, and model recycling to reduce training from scratch
How Model Atlas is built: a structured graph of models, attributes, and weight-space transformations connecting fine-tunes and lineages
Weight-space learning in practice: inferring functionality, performance, and lineage from weights when metadata is missing
What makes it scale: key methods behind repository-level analysis- and what the graph reveals beyond search and tags
This webinar is designed for ML engineers, applied researchers, and AI teams working with open model ecosystems.
Register now to save your spot.