MedGemma 1.5: Google's Open-Source Medical AI with 3D Imaging (Tutorial)

Run Google’s new MedGemma 1.5 locally for private, GPU‑accelerated medical AI. In this video, we break down what MedGemma 1.5 is, how its open‑weight 4B model runs on consumer hardware, and where it fits in the medical imaging and clinical text landscape. Using a hands‑on Google Colab notebook, we install MedGemma 1.5 from Hugging Face, benchmark it against proprietary trillion‑parameter models and GPT‑4 on medical QA, and demo real use cases like chest X‑ray interpretation, disease classification, anatomical localization, and medical document information extraction. --- Connect with me 🤗 - X: https://x.com/_alejandroao - LinkedIn: https://www.linkedin.com/in/alejandro-ao/ --- 🔗 Links - 📦 Code from this video (Colab notebook): https://github.com/huggingface/hub-tutorials/blob/main/notebooks/05-medgemma-1.5.ipynb - 📝 Written version: https://alejandro-ao.com - 🚀 My AI Roadmap: https://learn.alejandro-ao.com/ - ❤️ Buy me a coffee... or a beer (thanks!): https://buymeacoffee.com/alejandro.ao --- 🤓 Topics Covered - MedGemma 1.5 overview, benchmarks, and open‑weight licensing - Local, GPU‑accelerated MedGemma setup with Hugging Face - Medical imaging and document understanding demos --- ⏱️ Chapters 00:00 - Introduction 00:42 - What is MedGemma 1.5? 02:25 - Benchmarks & Performance 04:32 - Hands-on Demo