Apheris launched the ADMET Network, a federated data initiative that permits pharmaceutical partners to train joint models for absorption, distribution, metabolism, excretion and toxicity predictions without sharing raw proprietary data. Founding members include Lundbeck, Orion, Recursion and Servier, each committing substantial datasets to a governed training framework designed to reduce ADMET‑related clinical failures. Separately, Tamarind Bio raised $13.6 million Series A to commercialize an AI model coordination platform that democratizes access to biological machine‑learning models for bench scientists. Tamarind’s platform hosts >200 models and targets chemistry and protein design workflows. Together these moves indicate industry momentum toward shared, privacy‑preserving model training plus tools that lower the barrier for non‑computational biologists to run production‑grade models. Sector participants framed federated training and usable model infra as critical to reducing downstream attrition driven by ADMET liabilities.