Marius Varna
Software Engineer at Eneba. MSc in Software Systems Engineering (KTU). Full-stack - TypeScript/React, Symfony, FastAPI - with a research background in applied ML for medical imaging.
01 - Projects
Selected projects
Things I built and maintain.
Odis - Diagnostic Intelligence
On holdDermatological decision-support prototype. Prototype-based classifier (ProtoPNet) on HAM10000 covering 7 lesion categories - predictions ship with the visual prototypes behind them, so a clinician can audit the reasoning instead of trusting a score.
- Prototype-based classification with expert feature anchoring (ISIC 2018 attributes)
- Web demo for exploring cases and prototype evidence
timer - CLI countdown / stopwatch / pomodoro
ActiveCountdown, stopwatch, and pomodoro in one small Rust binary. Configurable via file, no runtime to install.
02 - Research
Research
Peer-reviewed work.
Interpretable Classification of Skin Lesions using ProtoPNet and Expert Feature Integration
AcceptedPrototype-based classifier on HAM10000, anchored to clinical features from ISIC 2018. Matches strong baselines on 7 lesion categories while exposing decision logic in terms dermatologists can review.
03 - Experience
Experience
Work and education.
Full-stack product work on production systems.
Internal backend systems with FastAPI: API design and implementation for research tooling.
Thesis: "Explainable Deep Learning System for Skin Lesion Classification" - grade 10/10.
Core stack
Frontend
- TypeScript
- React / Next.js
- Tailwind CSS
Backend
- Symfony (PHP)
- FastAPI (Python)
- PostgreSQL
- REST APIs
ML
- PyTorch
- OpenCV
- scikit-learn
04 - Contact
Contact
Open to interesting work and research collaboration.