Data Scientist, Clinical Outcomes

Midi Health•San Francisco, CA

1d•Hybrid

About The Position

Midi is seeking a pragmatic, hands-on Data Scientist to build and own the quantitative foundation for clinical care delivery, patient treatment plan refinement, and retention. The role connects clinical outcomes with business growth by using patient data to support personalized treatment paths, clinical efficacy, patient satisfaction, and long-term retention. The Data Scientist will work with structured clinical metrics and unstructured patient communications, including lab values, symptom severity scores, dosage trajectories, visit transcripts, ambient scribe outputs, and intake survey dialogue. Working across Data Science, Clinical Operations, and Product, this person will transform messy, multimodal patient data into clinical decision support and personalized care pathways. This is a full-time W2 position based in Palo Alto or San Francisco, with two days per week in the office. The role reports to the Director of Data Science + Analytics. The base salary range is $144,000–$172,000, depending on experience, plus equity and benefits.

Requirements

Advanced proficiency in SQL and strong Python skills for predictive modeling and data transformation.
Practical experience with clustering, regression or classification, and survival analysis, using tools such as scikit-learn, statsmodels, or lifelines.
Hands-on experience with NLP techniques and LLM frameworks, such as transformers or Claude API, to parse, extract, and classify unstructured text.
Working knowledge of quasi-experimental design, propensity score matching, and counterfactual reasoning with observational datasets.
Comfort working with messy, semi-structured, or missing healthcare data, including lab results, Patient-Reported Outcome Measures, and EHR notes.
A pragmatic 0-to-1 builder mindset, including willingness to do foundational data architecture, clean JSON outputs, structure features, and validate baseline models.
Ability to communicate technical modeling results as clear, intuitive insights for Clinical Operations and Product partners.
2 to 4 years of hands-on data science or analytics experience delivering high-impact solutions.
Experience working with healthcare, telehealth, or digital health datasets.
Demonstrated track record of converting unstructured text or transcript data into structured modeling pipelines.
Bachelor’s or Master’s degree in Biomedical Data Science, Health Informatics, Biostatistics, or a related quantitative field.
Must be authorized to work in the United States without current or future sponsorship needs.

Responsibilities

Build reliable modeling tables and feature stores by parsing, cleaning, and joining raw clinical datasets, including lab results, symptom surveys, medication history, and visit transcripts.
Design data-processing pipelines that convert unstructured clinical notes, orders, and visit transcripts into structured, queryable relational tables.
Develop lightweight NLP and LLM-based pipelines using methods such as entity extraction, sentiment analysis, and topic modeling to identify patient intent and preferences.
Detect unaddressed obstacles and early side-effect reports in patient dialogue before they contribute to treatment discontinuation.
Develop unsupervised models to cluster patients into lookalike cohorts using baseline labs, symptom profiles, and transcript-derived preferences.
Build predictive models to estimate symptom reduction, time-to-relief, and potential side-effect risks across alternative treatment paths.
Use observational study methods, including propensity score matching, difference-in-differences, and survival analysis, to evaluate non-randomized care pathways and estimate counterfactual outcomes.
Identify clinical touchpoints and dosage adjustment milestones that influence program adherence, and connect clinical milestone achievement with retention and lifetime value.
Partner with the clinical research team to support peer-reviewed publications, white papers, and payer studies.
Translate analytical findings, treatment protocol insights, and patient drop-off trends into actionable workflow improvements for clinical operations and care teams.