Maria Busygina / Machine Learning + Full-Stack / Dubai, UAE
// Selected Work 2024-2026
NEXT.JS / FASTAPI / SUPABASE / MOBILENETV2 / OPENCV
full-stack document scanner and classifier. a mobilenetv2 transfer-learning model at ~98% test accuracy, a five-pass detection cascade for hostile lighting, and a feedback-to-retraining loop, shipped as a deployed web app.
View Project ↗SIGLIP-2 / CROSS-ATTENTION / SHAP / MISTRAL 7B
capstone research system fusing image, text, and tabular evidence to flag fraudulent claims, with an explanation layer that shows its reasoning instead of hiding behind a score.
View Project ↗X3DH / DOUBLE RATCHET / FASTIFY / REACT NATIVE
end-to-end encrypted messenger built solo, from key agreement protocol to mobile client, running in production.
View Project ↗PYTHON / OPENCV / MEDIAPIPE
real-time american sign language recognition from a webcam feed, landmark tracking to live classification.
View Project ↗cs undergraduate (big data & ai) in dubai, self-funded through a freelance tutoring business i run end to end. i ship fast, learn stacks in days, and care about the difference between a demo and a product. currently: capstone research, this site, and plotting a move to europe.
// System Specs
Python / TypeScript / PyTorch-TF / React / Next.js / FastAPI / Supabase / OpenCV / GSAP
Status: Open to Internship / Part-Time / Contract