About

Building an open-source solution for semiautomatic terminology mapping in healthcare, powered by modern AI tools and grounded on the OMOP Common Data Model. Project funded by Precision Data and currently under development.

Leading language-modeling research for information extraction from clinical notes focused on breast cancer patient care. Investigating the potential of LLMs for multidisciplinary agentic tumor boards in complex patient scenarios. Building and evaluating information retrieval and clinical decision support pipelines.

Contributing physician for the OpenAI HealthBench benchmark, supporting advanced human data initiatives to enhance the safety of AI models. Primary responsibility was to rigorously evaluate the clinical accuracy and quality of AI-generated responses to health queries, directly contributing to the development of safer AI solutions.

Clinical practice as a family doctor in a private primary care center.

Led data science and quality improvement projects in the primary care centers, including patient scheduling optimization and reducing healthcare professionals' after-hours work.

Worked on ChartSmarter, an AI-powered application for electronic health records (EHR). Led backend and AI development, implementing voice transcription, EHR API integration, and automated information extraction. Designed and deployed LLM-based form completion, reducing clinician workload by transcribing and pre-filling EHR forms. The system was tailored for rehabilitation clinics in the U.S. using Kipu software.

Contributed to family doctors' training through lectures, clinical supervision, curriculum planning, and as a research advisor.

Clinical practice as a family doctor. Led and served as a technical reference for healthcare professionals of a private primary care center.

Contributed to the digital product team in developing the EHR and a system for automated patient coordination, enhancing longitudinal care.

Research focused on medical coding automation with deep learning. Explored information retrieval algorithms, LLM orchestration, and benchmarking. Experience with every experimental step — from data collection, annotation, and analysis to model training, evaluation, and reporting. Led international collaborations. Led to three publications, including one presented at the NeurIPS 2025 GenAI4Health workshop.

Hands-on experience with language modeling and computer vision, all implemented from scratch. Projects included pneumonia classification from chest X-rays using several vision architectures, and interpretable neural networks using Grad-CAM. Tutored a student in the following cohort with one-on-one sessions.