Data Scientist | Physician | AI Product Solutions
High-Performance AI & Data Architecture for Complex Systems.
Bridging 5+ years of acute care clinical experience with advanced machine learning, NLP, and data analytics. I combine medical precision with AI solutions to build robust, scalable systems for high-stakes environments.
An end-to-end ML pipeline trained on 4,000+ clinical records using Decision Trees, Naive Bayes, and neural networks. Interact with the widget to render dynamic mortality risk and survival timelines based on engineered clinical features.
Simulated ensemble output · Clause threshold >= 60 flags escalation
S(t) = 1 − R·(1 − e⁻ᵃᵗ) · R = modeled risk coefficient · CI 95% · SEER-selected cohort, t ∈ [0, 60] months
Outperforming raw positive node counts in Decision Tree root splits.
Range 1 – 60 · total surgical nodes sampled
Range 0 – 15 · capped at Nodes Examined
LNR = 2 / 15 = 0.133
Lymph Node Ratio = Nodes Positive / Nodes Examined
In our decision tree model, raw positive lymph node count was passed over in favor of the engineered Lymph Node Ratio (LNR) as the single most critical root decision boundary. An LNR ≤ 0.349 served as the primary gatekeeper for favorable survival outcomes.
Standardized Linear Regression Impact on Survival Months.
Physical dimensions (Tumor Size) show a negative impact on survival time (-0.04 months/mm). Click below to integrate molecular biology.
Three production-grade clinical AI systems. Explore the architecture, the clinical logic, and run a live demo inside each card.
A hybrid LLM and rule-based architecture creating deterministic, longitudinal patient records. Engineered with state persistence, temporal reasoning, and Pydantic schemas to eliminate clinical hallucinations.
A two-track pipeline where agentic orchestration meets deterministic rule modules.
A high-performance drug-drug interaction engine built on FastAPI. Leverages CTE-based self-joins and optimized set-based SQL queries for instant, single-roundtrip multi-drug analysis.
A thin FastAPI service carrying a relational interaction graph in PostgreSQL.
A stateful conversational AI built with LangGraph and Ollama. Features automated symptom extraction, dynamic urgency scoring, and multi-layer clinical safety guardrails.
A stateful graph with a local inference runtime and layered safety valves.
A continuous EKG of my clinical rotations, data and AI roles. Scroll to trace the rhythm from bedside to data science.
Scaled AI EMR processing across hospital networks covering millions of patients. Translated trial protocols into machine-readable logic, boosting speeds by over 5x.
Designed accessible digital healthcare platforms and formulated growth strategies based on health data trends.
Executed clinical trials and synthesized complex data in high-pressure ICU, ER, and Cardiology environments.
Delivered fast-paced emergency management and diagnostic care across operations.
A diagnostic audit of my academic and clinical record: MD, data science, biomedical engineering, and GCP. Printed out in real time.
Grab a pill. Every skill is draggable. The stack spans clinical standards (FHIR / ICD-10) through statistical modeling and modern cloud-native infra.
I enforce strict data integrity, build scalable architectures, and deploy AI that solves structural problems. Open to roles in data science, AI engineering and product roles across all data-intensive sectors. Secure routes below. No chatbots, just a direct line.