Skip to content
Dr. Ram KumarDiabetes & metabolic health

Medical AI

A structured course in artificial intelligence for healthcare: from the statistics behind every diagnostic test to machine learning, deep learning, clinical language models, retrieval-augmented generation and AI agents.

It is written for clinicians who want to understand and evaluate AI, and for developers and data scientists who want to build it responsibly in a clinical setting. Each lesson is tagged by level and focus, explains the idea in plain language, shows working Python code where it helps, and ends with an exercise using synthetic or openly licensed data. Every claim is sourced.

Start here: choose your pathGlossary

The curriculum

Ten tracks, taken in order or dipped into as needed. Later tracks list the earlier lessons they build on.

  1. Foundations

    What AI in healthcare is, the kinds of data it learns from, the standards that hold that data together, and the statistics needed to judge any model.

    6 lessons planned · 6 available

  2. Machine learning

    Classical machine learning on tabular clinical data: framing the question, building a cohort without leakage, and training, explaining and checking risk models.

    9 lessons planned · 9 available

  3. Deep learning & imaging

    Neural networks for images and signals: classification, segmentation, ECG and time-series models, and imaging foundation models.

    5 lessons planned · coming soon

  4. Clinical NLP

    Getting structure out of clinical text: notes, named-entity recognition, coding, embeddings and de-identification.

    4 lessons planned · coming soon

  5. Generative AI & LLMs

    How large language models work, how to prompt and adapt them for clinical tasks, and why hallucination and grounding matter more in medicine than anywhere else.

    7 lessons planned · 7 available

  6. Retrieval-augmented generation (RAG)

    Grounding LLM answers in trusted sources such as clinical guidelines: chunking, embeddings, retrieval, citations and evaluation.

    7 lessons planned · coming soon

  7. Agentic AI

    LLM agents that use tools and take steps: tool calling against FHIR data, multi-agent workflows, MCP, guardrails and human oversight.

    6 lessons planned · coming soon

  8. Evaluation, safety & deployment

    What happens after the model works on your test set: external validation, dataset shift, bias, reporting standards, red-teaming and monitoring.

    5 lessons planned · coming soon

  9. Regulation, ethics & governance

    How clinical AI is regulated and governed around the world: software as a medical device, the US, EU, UK and India, WHO guidance, liability and consent.

    7 lessons planned · coming soon

  10. Applied case studies: diabetes & metabolic health

    End-to-end projects on problems from diabetes care, bringing together the methods from every earlier track.

    5 lessons planned · coming soon

Latest lessons

Articles

Standalone pieces alongside the course.

Paper explainers

Important healthcare-AI studies, explained: what was done, what was found, and what it does and does not show.

Regulation & news

Changes in how clinical AI is regulated and governed around the world, dated and sourced.

Tool & model reviews

Vendor-neutral looks at models, libraries and tools used in healthcare AI. Mentions are not endorsements.

Practice perspectives

A practising physician's view on where AI helps, where it doesn't yet, and what to ask before using it.

Educational content only. This material teaches methods; it is not clinical advice. Code, models and examples here are not validated medical devices and must not be used to make decisions about real patients without proper clinical validation, governance and any required regulatory approval. Examples use synthetic or openly licensed de-identified data only. Never paste identifiable patient information into public AI tools. Mentions of models, tools or vendors are for identification, not endorsement.