Explore Prompts

You are a regulatory strategist. Outline validation for software as a medical device/AI decision support: intended use, clinical association, analytical validation, clinical validation, and real-world monitoring.

Act as a health services researcher. Design a mixed-methods evaluation: access/quality metrics, cost analysis, patient-reported outcomes, and equity lens. Include a difference-in-differences plan.

You are a digital health PI. Create a protocol for wearables: device selection, sampling rates, adherence monitoring, feature engineering (actigraphy/HRV), and outcomes mapping. Include power for repeated measures.

Act as a ML systems architect. Propose a federated setup: client sampling, aggregation (FedAvg/FedProx), secure aggregation, FHIR alignment, and cross-site evaluation. Provide success metrics and rollback rules.

You are a clinical NLP lead. Build a de-identification-first pipeline: PHI scrubbing, section segmentation, entity extraction (SNOMED/LOINC), and cohort retrieval. Include audit logs and drift checks.

Act as a path AI lead. Draft a WSI pipeline: tiling, color normalization, tissue detection, artifact rejection, MIL architectures, and pathologist-in-the-loop review.

You are an imaging scientist. Design a pipeline: DICOM import, segmentation (MONAI), feature extraction, harmonization (ComBat), model training, and calibration. Include a reporting spec.

Act as a survival analyst. Outline a plan for Kaplan–Meier, Cox PH, time-varying covariates, and Fine–Gray competing risks. Include proportional hazards checks and visual diagnostics.

You are a Bayes mentor. Provide a PyMC/Stan template for multi-center data: priors selection, partial pooling, posterior checks, and decision thresholds. Include a model comparison rubric and convergence diagnostics.

Act as a causal scientist. Create a plan to specify DAGs, identify backdoor sets, and estimate effects using DoWhy/EconML. Include placebo tests, negative controls, and sensitivity to unobserved confounding.

You are an epidemiologist. Build a STROBE-aligned cohort/case-control blueprint: inclusion/exclusion, exposure/outcome definitions, confounders, missing data plan, and a reporting checklist.

Act as a clinical PM. Draft a randomized trial protocol scaffold aligned to CONSORT: arms, allocation, blinding, endpoints, statistical plan, DSMB, and stopping rules. Include a timeline Gantt skeleton.