478K+
clinical documents processed in a single de-identification benchmark
Clinical NLP · Healthcare Data · Medical Coding Systems
Hi, I'm Ozgur Caglayan
Models Hub Team Lead at John Snow Labs · Gaziantep, Türkiye
I take messy, complex clinical information and turn it into something clear, usable, and meaningful — building data pipelines and ML/NLP systems with Python, SQL, and Databricks. I'm especially drawn to clinical coding systems like SNOMED CT and ICD, and how we make medical data speak the same language across systems.
Find me on
Proof in numbers
Measurable results from real pipelines
11.7M+
tokens analyzed across different NER pipeline architectures
~20x
CPU runtime speedup achieved (9.4 hours → 26 minutes)
What I do
End-to-end healthcare NLP, from raw clinical text to standardized codes
Clinical NLP & De-Identification
I design NER-based pipelines to detect and remove PHI from clinical text — rule-augmented, hybrid, and zero-shot approaches.
Medical Coding & Terminology Mapping
I build autocoding and entity-resolution systems for ICD and SNOMED CT, mapping clinical language onto standardized codes.
Healthcare Data Standardization
I build data pipelines and ML systems on Python, SQL, and Databricks so structured and unstructured healthcare data can move across systems reliably.
Tech stack: Python, SQL, Databricks, Spark NLP, AWS
Let's start a projectCareer timeline
Five years of turning raw data into usable systems, now focused on healthcare
Models Hub Team Lead
Feb 2026 – PresentJohn Snow Labs
Collaborate with cross-functional teams to translate healthcare data into production-ready solutions. Build and maintain scalable data pipelines with Python, SQL, and Databricks for large-scale structured and unstructured healthcare data, and develop ML/NLP workflows for extracting insights from clinical datasets.
Data Scientist
Nov 2025 – Feb 2026John Snow Labs
Worked on clinical coding and interoperability challenges — SNOMED CT, ICD mapping, and structured medical data representation — improving data quality and consistency for downstream analytics and ML.
Data Scientist — Python | Web Scraping | Automation | ML | DL
Mar 2020 – Sep 2025Self-Employed
Delivered freelance data and automation projects end to end — Python tools for data processing and scraping, lightweight applications and backend logic, and data pipelines across client systems, turning loosely defined ideas into working implementations.
Data Scientist
Dec 2018 – Mar 2020yDataLabs
Built end-to-end data workflows in Python and SQL, with scraping systems to gather and structure large volumes of web data, then applied ML to explore patterns and prepare training data for iterative model work.
Case studies
Problem, approach, result
De-Identification
Clinical De-Identification at Scale
- Problem
- Removing PHI from clinical text without sacrificing speed or accuracy.
- Approach
- Benchmarked rule-augmented NER, hybrid NER + zero-shot, and fully zero-shot pipelines across Databricks-AWS (CPU/GPU) and Google Colab.
- Result
- Cut CPU runtime from 9.4 hours to 26 minutes on 478K documents / 11.7M tokens; GPU acceleration brought the mid-tier pipeline to ~1 hour and the zero-shot pipeline to 6.5 minutes.
Healthcare Interoperability
From Doctor Notes to Structured Data
- Problem
- It's unclear how code creators, code mergers, and application layers fit together to make healthcare data usable.
- Approach
- Mapped the three-layer ecosystem: code creators (WHO, SNOMED International), code mergers (UMLS, OHDSI), and application layers (FHIR APIs, OMOP).
- Result
- A reference framework that helps healthcare NLP teams reason about terminology-integration decisions.
Open Source
John Snow Labs NLP Ecosystem
- Focus
- Contributing to and building on top of Spark NLP for Healthcare, plus a dedicated Medical-Code-Systems repository.
- Where
- Public work across spark-nlp, spark-nlp-workshop, and johnsnowlabs — the gateway into the John Snow Labs ecosystem.
About
I'm a Data Scientist working in the healthcare space, mainly around clinical NLP and structured medical data. I enjoy taking messy, complex clinical information and turning it into something clear, usable, and meaningful. A big part of my interest is clinical coding systems like SNOMED CT and ICD, and how we can make medical data talk the same language across systems. For me, the most exciting part is working on problems where data actually connects to real human impact.
Certifications & Community
- AWS Partner: Technical Accredited — Training Badge
- Contributor, John Snow Labs open-source NLP libraries (Spark NLP, Spark NLP Workshop)
- Author, clinical NLP writing on Medium
Résumé
Full experience, skills & certifications — PDF
Let's discuss opportunities
Reach out for collaboration, consulting, or hiring inquiries