The 'Med AI' Capsule Podcast by Dr Avneesh Khare

In this episode of The Med AI Capsule, we break down federated learning through a simple, conversational walkthrough of how hospitals can collaboratively train AI models without ever sharing patient data. The discussion covers why this matters for privacy, how the workflow actually operates in practice, the safeguards needed to prevent information leakage, and the real-world limitations that still require clinical oversight. It’s a clear, practical primer on the idea of collaboration without data sharing—and why that matters for building trustworthy medical AI.
Note: This audio podcast is an extension of The Med AI Capsule newsletter by Dr Avneesh Khare. You can explore more at www.avneeshkhare.com.
Disclaimer: This audio podcast was generated using AI tools and is based on content from the related newsletter issue. While all the information is curated from reliable sources, it is only intended for educational and informational purposes. AI LLMs can make mistakes - please check all the important information at your end. For any issues, please reach out at [email protected].


This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit avneeshkhare.substack.com

In this episode of The Med AI Capsule, we decode RAG—how it connects AI to evidence. Instead of relying on training data alone, RAG retrieves current clinical guidelines and research before answering. We explore why this matters for clinical reliability and what still requires human judgment.
Note: This audio podcast is an extension of The Med AI Capsule newsletter by Dr Avneesh Khare. You can explore more at www.avneeshkhare.com.
Disclaimer: This audio podcast was generated using AI tools and is based on content from the related newsletter issue. While all the information is curated from reliable sources, it is only intended for educational and informational purposes. AI LLMs can make mistakes - please check all the important information at your end. For any issues, please reach out at [email protected].


This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit avneeshkhare.substack.com

In this episode of The Med AI Capsule, we look at how medicine is racing ahead and why explainable AI has become the central requirement for safe adoption. We unpack how tools like heat maps, SHAP scores, and rule extraction pull back the curtain on complex models and why these explanations are only clues about the model’s logic, not biological truth. The goal is simple: build systems clinicians can trust without surrendering judgment.
We then move through the latest research and deployments. Speech-based models for early dementia show strong accuracy, but clinical readiness remains low because workflows are ignored. On the ground, startups like Vuha Med Data are scaling cervical cancer screening in low-resource environments, while AI tools are already improving safety, communication, and clinician wellbeing through infection prediction, patient-facing chat systems, and ambient AI scribes that cut burnout within weeks.
We close with the policy and professional shifts that this rapid growth demands. New AMA requirements and India’s DPDP Act are reshaping training and data governance. And specialties like anesthesiology face an uncomfortable but necessary question: if machines handle the mechanics, how should clinicians redefine their role? The episode ends on the core challenge that will shape the decade ahead: using AI’s speed without losing the human connection that defines care.
Note: This audio podcast is an extension of The Med AI Capsule newsletter by Dr Avneesh Khare. You can explore more at www.avneeshkhare.com.
Disclaimer: This audio podcast was generated using Google NotebookLM and is based on content from the related newsletter issue. While all the information is curated from reliable sources, it is only intended for educational and informational purposes. AI LLMs can make mistakes. Please check all the important information. For any issues or inaccuracies, please reach out at [email protected].


This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit avneeshkhare.substack.com

In this episode of The Med AI Capsule, we explore how AI is moving from pilot to practice—redefining diagnostics, clinical workflows, and critical care.
We begin with OpenAI’s clinical copilot, which cut diagnostic errors in a study with Penda Health and showed real-world gains in patient safety and clinician trust. Next, we spotlight GE HealthCare’s dominance in FDA approvals and Mayo Clinic’s bold generative AI deployments powered by NVIDIA’s Blackwell infrastructure.
We also take a closer look at Cloudphysician’s Smart-ICU model, where AI copilots and video analytics are transforming critical care in many Indian hospitals. On the research front, we cover a machine learning model that enables interpretable, personalized antibiotic therapy—offering a practical path to combat resistance.
Plus, we discuss how digital patient systems powered by large language models are elevating medical training and whether ChatGPT can match emergency physicians in structured scenarios. Tune in to understand how precision, explainability, and real-world deployment are shaping the AI-powered future of medicine.
Note: This audio podcast is an extension of The Med AI Capsule newsletter by Dr Avneesh Khare. You can explore more at www.medaicapsule.in.
Disclaimer: This audio podcast was generated using Google NotebookLM and is based on content from the related newsletter issue. While all the information is curated from reliable sources, it is only intended for educational and informational purposes. AI LLMs can make mistakes. Please check all the important information. For any issues or inaccuracies, please reach out at [email protected].


This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit avneeshkhare.substack.com

In this episode of The Med AI Capsule, we dive into how AI is transforming diagnosis, medical education, and specialist workflows. We look at Microsoft’s new MAI-DxO benchmark, designed to evaluate clinical reasoning under cost constraints—offering a fresh lens on diagnostic AI in high-stakes settings.
We also cover Mayo Clinic’s StateViewer, an AI tool that classifies nine dementia types with a single scan, and explore how China’s Guangdong Medical University is pioneering AI medical education through immersive, large-model-powered platforms.
Plus, we discuss new studies on interpretable AI for Crohn’s disease detection, and how extracurricular education can bridge the AI literacy gap for pre-med students. Tune in for insight into how diagnostic precision, educational reform, and transparent AI design are shaping tomorrow’s care.
Note: This audio podcast is an extension of The Med AI Capsule newsletter by Dr Avneesh Khare. You can explore more at www.medaicapsule.in.
Disclaimer: This audio podcast was generated using Google NotebookLM and is based on content from the related newsletter issue. While all the information is curated from reliable sources, it is only intended for educational and informational purposes. AI LLMs can make mistakes. Please check all the important information. For any issues or inaccuracies, please reach out at [email protected].


This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit avneeshkhare.substack.com