Field notes on healthcare AI

Working with clinical data

"The hardest part isn't the model — it's cleaning the data." We heard this in 4 separate interviews. Clinical data is noisy, anonymization is slow, and regulation keeps shifting.

Where are the jobs?

• Imaging AI: Mature in radiology and pathology.
• Clinical decision support: Systems that suggest, never decide for, the doctor.
• Operational AI: Patient flow, scheduling, resource planning.

What to learn first?

PyTorch and basic statistics are enough. Compliance topics like HIPAA/KVKK are learned on the job.