Joseph Mifsud

I’m a Data Scientist with 5+ years of experience building machine learning models, statistical analyses, and production data pipelines for healthcare and commercial applications. I focus on reliable data systems that support reproducible research, scalable ML workflows, and robust analytical solutions.

Most recently, as Founder at Kaleh LLC and Trace Data, I built a production job market dataset from 9,800+ structured sources, ingesting 13,000 new postings daily for analytical and ML applications. I designed deterministic pipelines to normalize heterogeneous records, preserve source provenance, and create reproducible datasets, including data quality workflows with schema validation, canonicalization, and SHA-256 deduplication. I also built searchable infrastructure over millions of records using an inverted index and FastAPI, owning end-to-end ingestion, storage, API delivery, and production monitoring.

Previously at Shirley Ryan AbilityLab, I developed computer vision models for Parkinson’s disease detection from clinical videos, achieving F1 > 0.90 with rotation-invariant pose-based features. I built and deployed end-to-end ML pipelines for wearable sensor time-series and created PostgreSQL databases and ETL workflows for longitudinal clinical datasets, including migrating large-scale video data to AWS S3.