Today I had the pleasure of giving the keynote "𝙁𝙧𝙤𝙢 𝙄𝙢𝙖𝙜𝙚 𝙍𝙚𝙩𝙧𝙞𝙚𝙫𝙖𝙡 𝙩𝙤 𝘾𝙡𝙞𝙣𝙞𝙘𝙖𝙡 𝙍𝙚𝙖𝙨𝙤𝙣𝙞𝙣𝙜: 𝙈𝙪𝙡𝙩𝙞𝙢𝙤𝙙𝙖𝙡 𝙁𝙤𝙪𝙣𝙙𝙖𝙩𝙞𝙤𝙣 𝙈𝙤𝙙𝙚𝙡𝙨, 𝙎𝙝𝙖𝙧𝙚𝙙 𝘽𝙚𝙣𝙘𝙝𝙢𝙖𝙧𝙠𝙨, 𝙖𝙣𝙙 𝘾𝙡𝙞𝙣𝙞𝙘𝙖𝙡𝙡𝙮 𝘼𝙡𝙞𝙜𝙣𝙚𝙙 𝙀𝙫𝙖𝙡𝙪𝙖𝙩𝙞𝙤𝙣 𝙞𝙣 𝙍𝙖𝙙𝙞𝙤𝙡𝙤𝙜𝙮 𝙖𝙣𝙙 𝘿𝙚𝙧𝙢𝙖𝙩𝙤𝙡𝙤𝙜𝙮" at the Workshop on Large Foundation Models in Biology and Biomedicine (#LFMBio) @ #WACV2026. I'm happy to share the abstract and slides from the talk for those interested in this topic: lnkd.in/ebNijeWG Many thanks to @therealthapa, Suchi Bhandarkar, and all the organizers for the invitation! Wishing you continued success in future editions! 𝐀𝐛𝐬𝐭𝐫𝐚𝐜𝐭: Advances in multimodal foundation and vision-language models have accelerated progress in medical AI. Yet a central challenge remains: how can we move from image-level understanding to clinically grounded reasoning that is reliable and measurable? In this talk, I will present some of our recent research on 3D image retrieval, structured report generation, multimodal visual question answering (VQA), and clinically aligned evaluation across radiology and dermatology. In volumetric medical image retrieval, I will introduce 3D-MIR, the first benchmark for 3D medical image retrieval, and methods for adapting multimodal foundation models to volumetric clinical data. I will discuss supervised multimodal adaptation and comparative analyses across foundation models, including MedImageInsight, highlighting how representation learning and aggregation strategies influence lesion detection, characterization, and case similarity, with validation by clinical experts. I will also showcase practical applications from the Image Search Series, including automated 3D retrieval pipelines and retrieval-augmented systems for radiology and dermatology. The second and third parts of the talk will focus on structured radiology report generation and multilingual VQA in dermatology and wound care and will showcase how vision-language models can support multimodal clinical reasoning across images and text. In particular, I will highlight newly created datasets, including DermaVQA and WoundcareVQA, and their applications in AI for dermatology and wound care. The last part of the talk will present insights derived from community challenges such as recent MEDIQA shared tasks on dermatology and wound care VQA, and our latest initiatives: MEDIQA-EVAL 2026, which examines the alignment between automatic metrics and expert assessment in multimodal and multilingual VQA, and MEDIQA-CORE 2026, which focuses on multimodal reasoning and report reconciliation in radiology. #MedicalAI #HealthcareAI #VLMs #Benchmarks #MultimodalAI lnkd.in/ebNijeWG