Michael Hagar — Senior Data Scientist

PathAI

Senior Data Scientist, AI Pathology & LLM Products

Jul 2024 – Present

PythonSQLRAGLangChainLLM EvaluationSnowflake

Lead data science for AI pathology products, evaluating model reliability and clinical usability across cohorts, subgroups, annotation patterns, clinical metadata, edge cases, and model outputs.
Partner with pathologists, machine learning engineers, and product managers to translate model findings into release criteria, documentation, and practical improvements for clinical workflows.

Build RAG-based tools that help product, science, and clinical teams retrieve validated answers from model documentation, annotation guides, study materials, and validation results.
Design offline and production LLM evaluation checks for retrieval relevance, answer faithfulness, citation quality, hallucination risk, and safety guardrails to support reliable product decisions.
Create reusable evaluation datasets and error taxonomies that make model comparisons, regression testing, and release-readiness reviews more consistent and traceable.

Develop reusable Python and SQL workflows for dataset review, cohort analysis, slice-based performance assessment, edge-case investigation, validation, and ongoing model monitoring.

Senior Data Scientist, Real-World Evidence & Clinical AI

Oct 2021 – Jun 2024

PythonSQLNLPEmbeddingsSnowflakeReal-World Evidence

Led real-world evidence analyses across clinical, molecular, and outcomes data for oncology, precision medicine, study feasibility, treatment patterns, and outcomes research.
Built reusable cohort logic and analytical datasets for patient selection, treatment-pattern analysis, endpoint definition, variable derivation, and reporting.

Developed NLP- and LLM-assisted tools that helped teams query study criteria, data dictionaries, cohort definitions, analytical outputs, and supporting documentation.
Applied NLP, embeddings, and semantic search to identify and connect clinical concepts across structured data, medical terminology, and unstructured text.

Worked with clinicians, product managers, and data engineers to validate assumptions, communicate study limitations, and create reproducible QA and reporting workflows.
Established quality checks for cohort construction, variable derivation, missing data, and analytical outputs before findings reached clinical and business stakeholders.
Synthesized clinical and analytical findings into decision-ready narratives for study planning, stakeholder review, and downstream product requirements.

Data Scientist, Risk, Growth & AI Analytics

Apr 2018 – Sep 2021

PythonSQLXGBoostSnowflakeAirflowLooker

Owned analytics for fraud risk, onboarding, activation, customer growth, retention, and product performance across fintech workflows.
Developed and evaluated risk models to identify suspicious account behavior, transaction anomalies, payment issues, and cases requiring manual review.
Reviewed model thresholds and segment-level error patterns to balance detection quality, review capacity, customer experience, and operational risk.
Designed and analyzed experiments for onboarding flows, lifecycle campaigns, pricing updates, activation programs, and retention initiatives, from hypothesis through launch recommendation.

Prototyped internal search and chatbot tools that helped support and risk teams navigate policies, operational notes, and recurring issue patterns.

Built reusable SQL models and dashboards for funnel health, customer segments, risk queues, fraud trends, operational capacity, and growth reporting.
Documented model assumptions, monitoring signals, and escalation paths so risk and operations teams could act consistently on analytical outputs.
Partnered with product, risk, compliance, and operations teams to define requirements, prioritize investigations, and turn analysis into operational decisions.