Hello OHDSI Community,
I joined OHDSI less than two weeks ago and am looking to connect and collaborate with fellow implementers, working groups, and open-source contributors across the community.
Current Technical Focus & OHDSI Exploration
I am currently building R&D and proof-of-concept projects focusing on data pipelines, biomedical RAG architectures, and NLP:
OHDSI Tooling & OMOP CDM: Hands-on with OMOP CDM v5.4 and actively exploring the standard OHDSI ETL toolkit (WhiteRabbit, Rabbit-in-a-Hat, Usagi, Data Quality Dashboard) using the Synthea dataset.
Data Engineering & Architecture: Applying modern data pipeline designs—such as the Medallion Framework (Bronze/Silver/Gold) — to bridge raw source data preprocessing, cleaning, transformation, and analytics-ready datasets for machine learning.
Tech Stack: Core focus in Python (pipelines, ML/NLP algorithms) alongside background experience in Java and continuous integration/deployment (CI/CD/CT). I am also currently picking up my R skills to better align with the HADES and OHDSI analytics ecosystem.
Biomedical NLP & LLMs: Finalizing a prototype Biomedical Research RAG system (using Python, Milvus standalone vector DB, and Ollama/embeddings) and experimenting with medical text extraction models using domain-specific transformers like PubMedBERT.
Looking to Collaborate
I’m eager and ready to contribute to active community efforts, learn from experienced implementers, and lend a hand on technical challenges. I am particularly interested in connecting with individuals or teams working on:
- ETL & Data Pipeline Optimization (from source system transformations to OMOP CDM).
- NLP & Text Processing for unstructured clinical records.
- MLOps / Infrastructure for open-source observational health analytics tools.
If you are working on related open-source projects, leading a working group, or simply want to exchange ideas on solution architecture in healthcare, I’d love to connect and chat!
Looking forward to engaging with you all.
Best regards,
Thabo
My Philosophy: Learn-Apply: repeat