About

I build machine learning systems for medical devices, and I study why medical AI fails in ways standard benchmarks miss.

At Medtronic Surgical Innovation, I’m a Senior Principal R&D Engineer and company-wide Technical Fellow. I advise senior leadership on machine learning and data science strategy, and my day-to-day work spans predictive clinical models, generative AI, retrieval-augmented generation (RAG) systems for R&D knowledge search, and cloud data pipelines for clinical evidence generation. Before the machine learning years I spent a long stretch in engineering data management and product lifecycle systems, which is where I learned to care about data integrity and traceability. All told, that’s 19 years of engineering.

I’m also a Ph.D. candidate at the Secure and Assured Intelligent Learning Lab (SAIL Lab) at the University of New Haven, advised by Dr. Vahid Behzadan, with the degree expected in August 2026. My dissertation asks an uncomfortable question: if you reword a clinical question without changing its meaning, why does a medical Vision-Language Model (VLM) change its diagnosis? I built PSF-Med, a benchmark of 92,856 audited question-paraphrase pairs, to measure the problem, then used sparse-autoencoder interpretability to trace where it comes from and targeted fine-tuning to reduce it. The research page has the full story and the numbers.

Along the way I’ve published papers spanning medical AI safety, precision oncology, and medical device reliability, and co-invented two patents on machine learning for surgical devices. I review for ICLR, MICCAI, Machine Learning for Healthcare (MLHC), and IEEE ICMLA.

Education

I also hold a Deep Learning Nanodegree from Udacity (2017), the Tableau Desktop Specialist certification, and a Certificate in Project Management.