I have several retrospective ICU studies underway and I need an analyst who already feels at home inside both MIMIC-III and MIMIC-IV on PhysioNet. The work centres on pulling raw tables, stitching them together through patient and hospital-stay identifiers, and transforming that sprawl into tidy, analysis-ready cohorts. Every step—from initial SQL queries to the last model diagnostic—must be reproducible and clinically defensible. Day to day you will: • Extract rows from the core Meds, diagnoses, procedures, labs, and vitals tables, then link them reliably across encounters. • Clean and preprocess the high-dimensional data, flagging implausible values and resolving unit or time-stamp issues. • Derive clinically meaningful variables (severity scores, medication exposure windows, composite outcomes, etc.). • Run univariate and multivariable analyses in R or Python, iterating as study questions shift. • Document assumptions, address potential confounding and bias, and draft short interpretive notes that fold easily into manuscripts. What matters most is hands-on mastery of data extraction, table linking, and general database management within MIMIC. Solid grounding in observational study design, epidemiology, and EHR quirks is essential; a background in medicine or public health will make communication smoother. Working code in SQL plus either tidyverse/R or pandas/pySpark is expected. The immediate deliverable is a fully cleaned analytic dataset with the accompanying scripts and an outline of the statistical approach. After that, I plan to keep the collaboration open for additional projects and sensitivity analyses as new questions arise.