Harvard CS197: AI Research Experiences is a free, publicly available course book that walks readers through how to actually do AI research, not just how to use AI tools. Written by Pranav Rajpurkar, an Assistant Professor of Biomedical Informatics at Harvard Medical School, it distills a full semester of coursework — over 250 pages across 21 lectures — into a practical, self-paced guide.
The book covers a wide span of skills needed for modern AI research: working with language models and fine-tuning them, PyTorch fundamentals, Vision Transformers, cloud and GPU-based training workflows, and diffusion model fine-tuning, alongside the software engineering habits (Git, environment management, debugging) that make research code reproducible.
Beyond the technical stack, it places strong emphasis on the research process itself — reading and searching academic literature, generating and scoping research ideas, tracking experiments, running statistical comparisons between models, and writing and presenting findings clearly. A capstone project applies these skills to a medical AI case study.
This resource is best suited for students, early-career researchers, and self-taught practitioners who already have some programming background and want a structured, hands-on introduction to how AI research is actually conducted, from first experiment to finished paper.