Hi! I'm Sara, welcome to my little coding biosphere.

"Natura non facit saltus." — Linnaeus  ·  _{nature does not make leaps}

· Carol Davila University of Medicine

Second-year medical student with a parallel obsession with code, specifically, making deep learning useful in clinical and biological contexts. I care most about problems where the biology is genuinely hard and the data is genuinely messy.

I am drawn to cardiology and oncology as future specialties: two fields where programming and medicine are converging fastest.

• location → Bucharest, Romania 🇷🇴
• languages → Romanian · English · Français · Deutsch

Research Intern — Laboratory of Ultrastructural Pathology
“Victor Babeș” National Institute of Pathology, Bucharest · May 2026 — present

I work on the machine-learning side of the lab's AI in pathology programme, developing reproducible deep-learning workflows for digital histopathology together with the lab's pathologists.

My current work is centred on colorectal histology semantic segmentation — teaching computer-vision models to localise and distinguish pathological tissue patterns at pixel level.

What that currently involves:

• designing and training multi-class semantic-segmentation pipelines in PyTorch / Ultralytics
• group-aware cross-validation and held-out evaluation
• class-imbalance handling and error analysis
• external-generalisation testing across public and institutional data
• quantitative analysis of failure modes, including ordinal grading errors
• coordinating annotation and reference-standard criteria with pathologists
• building reproducible experiment, evaluation and reporting workflows

The broader aim is to move beyond “can the model segment this image?” towards:

what pathological information is the model learning, where does it fail, and can that behaviour generalise to real clinical material?

Current methodological interests within the project include ordinal-aware grading, boundary-sensitive segmentation, robust evaluation and domain shift.

_{Laboratory of Ultrastructural Pathology, IVB}

• Deep learning for medical imaging, pathology & multi-omics
• Computational Medicine
• Cancer genomics & oncoproteomics
• CRISPR guide RNA design & off-target prediction
• Microbial classification
• Cardiac electrophysiology (aspiring)

I'm an introvert who recharges in forests and thinks best with a stylus in hand. What I love doing most, when I'm not swamped with exams:

• embroidery
• drawing — both digital and traditional; mostly anatomy and odd, funny little creature designs - also, ponies! hardcore MLP fan here!
• reading high-fantasy books and re-analysing the political dynamics between the elves of Mirkwood and the dwarves of Erebor.
• video games — player and hobbyist game dev; HTML5 canvas and R are my playground; Horizon Zero Dawn & Assassin's Creed enjoyer
• writing — I keep a blog: neural-decay, which is my digital diary of thoughts and observations
• catching up on new articles on cellular biology and computational biology
• people watching & daydreaming
• tinkering & fixing my jewerly
• playing with my Zeiss microscope