Healthcare data scientist working in causal inference, quasi-experimental program evaluation, and applied ML. Four years evaluating pharmacist-led clinical programs at a 60,000-patient managed care system — most recently a PSM + DiD evaluation that documented $83.50 PMPM cost reduction (p = 0.0027) across a COPD cohort.

M.S. Data Science, UC San Diego (starting Fall 2026) · B.S. Mathematics, UC Irvine

Available: Full-time · Open to contract / full-time · Remote-friendly

📧 Samir2000VIP@gmail.com · LinkedIn · Irvine, CA

Data Scientist across 11 clinics. Methods chosen to defend causal claims a reviewer would actually question.

• COPD program evaluation (n = 997, PSM + DiD): $83.50 PMPM cost reduction (p = 0.0027), driven by lower ED (p = 0.002), inpatient (p = 0.04), and readmission (p = 0.002) utilization.
• Post-discharge pharmacist intervention (n = 878, negative binomial regression): 22% reduction in 30-day readmissions (IRR = 0.78, p = 0.02).
• Heart failure outcomes manuscript — under peer review.
• AFib anticoagulation care-gap analysis — poster, ASHP National Conference.

Five-method stack (PSM, IPW, AIPW, Double ML, Causal Forest) built on DoWhy's identify–estimate–refute workflow. Placebo and random-common-cause refuters at each stage; method agreement used as a robustness check rather than a single point estimate. Why it matters: demonstrates the full causal pipeline most healthcare ML projects skip.

XGBoost / LightGBM ensemble over 1.69M ClinVar variants. Leakage-corrected AUC 0.985, macro-F1 0.948. Feature ablation defends against gene-name memorization: consequence + LoF alone reach AUC 0.97, while gene-only collapses to 0.78. SHAP per-variant audit; Llama 3 / Claude narrative engine generates ACMG/AMP-style reports. Why it matters: pre-empts the leakage critique a reviewer would lead with.

• ClinicalRAG — RAG over 220 clinical documents with retrieval and refusal as first-class metrics: 97.6% condition recall, 85.7% citation rate, 95.2% abstention accuracy.
• Diabetic Retinopathy — Custom CNN for 5-class DR grading. Weighted F1 = 0.94, outperforming ResNet-50 and VGG-16 on the same split. Grad-CAM confirms attention to clinically meaningful pathology. Paper.
• REIGN — Cross-era NBA impact models over 29,969 player-seasons with era-specific z-score normalization.

Causal & Statistics — PSM, DiD, IPW, AIPW, Double ML, Causal Forest, negative binomial & other GLMs, survival analysis ML — XGBoost, LightGBM, scikit-learn, TensorFlow / Keras, SHAP LLM / NLP — RAG, LangChain, ChromaDB, HuggingFace Transformers Healthcare data — EHR, claims, pharmacy, ICD-10, HCC, PMPM Languages & delivery — Python, SQL, R, Power BI, FastAPI, Git

_{Outside work: 2500+ rated chess · basketball · piano}