Suggested paths
Time-boxed routes through the roadmap, built only from resources in the list. Each step links to the resources it names.
7-day "AI-literate clinician" (≈8–10 h)
- Day 1
YUVA AI for ALL, modules 1–3 (#3).
- Day 2
YUVA modules 4–6 plus certificate.
- Day 3
Google Skills Generative AI for Healthcare plus lab (#1).
- Day 4
AMA ChangeMedEd modules: Intro and Methodologies (#5).
- Day 5
AMA Practical Applications and Precision Health modules.
- Day 6
Read the ICMR AI ethical guidelines summary sections (#39) and skim the DPDP Rules phasing (#40).
- Day 7
Read the TRIPOD+AI checklist (#22) against one AI paper from your specialty.
30-day "Use and Evaluate" (≈4–5 h/week)
- Week 1
the 7-day path.
- Week 2
Stanford Introduction to Clinical Data (#13) and Fundamentals of ML for Healthcare (#43).
- Week 3
Stanford Evaluations of AI Applications (#21; claim 9.50 CME), plus PROBAST+AI (#23) and STARD-AI (#24).
- Evaluations of AI Applications in Healthcare #21
- PROBAST+AI: an updated quality, risk of bias, and applicability assessment tool for prediction models using regression or artificial intelligence methods #191
- The STARD-AI reporting guideline for diagnostic accuracy studies using artificial intelligence #192
- Week 4
FDA CDS guidance (29 Jan 2026 version) and webinar slides (#33–34), FDA PCCP guidance (#31), and the CDSCO MDSW guidance (#37). Write a one-page "Is this tool a device in the US and India?" memo for a tool you use.
90-day "Build" (≈6–8 h/week, some Python)
- Weeks 1–4
the 30-day path, plus Kaggle Learn Python, Intro to ML and Intro to Deep Learning (#18), and Google MLCC (#56).
- Weeks 5–8
DeepLearning.AI AI for Medical Diagnosis and Prognosis (#46–47). Begin PhysioNet credentialing (CITI training) for MIMIC-IV (#15).
- Weeks 9–10
MONAI tutorials on segmentation (#54), plus CheXpert Plus (#16) or TCIA (#17).
- Weeks 11–12
Hugging Face Agents Course Unit 1 certificate (#64); DeepLearning.AI A2A healthcare multi-agent build (#60); the MCP course Unit 1 (#65).
- Capstone
a MIMIC-based risk model reported against TRIPOD+AI, or a human-in-the-loop documentation agent. Evaluate it with the Harvard AIM II evaluation framing (#73).