Ruslan

Ruslan Dovhai Education

BSc in IT & Business Analytics 2024–2027
Ukrainian Catholic University (UCU), 3rd year
⋄ Algorithms & Data Structures; Machine Learning; Mathematical
Methods of Machine Learning; Big Data Technologies; Computer Vis-
ion; Econometrics; Financial Analysis; Linear Algebra; OOP (Java);
Programming Fundamentals (Python); Probability & Statistics.

Certificate
⋄ Mathematical Methods in Applied Problems — UCU Faculty of
Applied Sciences, 180 hrs (6 ECTS), Feb–May 2025. Mentored by
Prof. A. Iosevich (University of Rochester, USA). Topics: optimal
transport theory, Fourier analysis, numerical optimisation, scientific
research methodology.
Profile
GitHub · Report · Certificate
4rd year IT & Business Analytics student
at Ukrainian Catholic University with
strong mathematics background and Selected Projects
hands-on experience in ML/DL, data
analysis, and software engineering. Built
Chess Web App (React, Java Spring, PostgreSQL, Redis). 2025
end-to-end pipelines across computer
⋄ Independently designed and built a full-stack chess platform from
vision, time series, and forecasting. Writes
scratch, without tutorials or AI assistance. Implemented complete
clean, reliable code and is comfortable
piece logic using bitboards for efficient move generation; designed
moving across data, models, and services.
and documented REST API with Java Spring; built React SPA front-
Passionate about solving complex
end with clean UI; Redis session caching; PostgreSQL persistence;
problems. Interested in Data Science and
FEN-style game snapshots for state management.
quantitative research.
GitHub
Contact details

@ [відкрити контакти](див. вище в блоці «контактна інформація») Real-time Hand Gesture Recognition (Python, OpenCV, Me-
H [відкрити контакти](див. вище в блоці «контактна інформація») diaPipe). 2025
m github.com/Ruslan4040 ⋄ Team project (3 members). Built a real-time gesture recognition
m [відкрити контакти](див. вище в блоці «контактна інформація») pipeline using MediaPipe and a custom labeled dataset with full data
preprocessing and feature extraction. Derived full SVM from first
principles: Lagrangian, KKT conditions, dual formulation with RBF
Skills
kernel, projected-gradient solver; one-vs-rest multi-class training.
• ML, Deep Learning, CV, Generative AI Achieved stable classification in live camera input; ≈0.95 accuracy
on held-out tests.
• Supervised and Unsupervised Learning
• Classification, Regression, Clustering, GitHub · Report · Presentation
Dimensionality Reduction
• SVM, Kernel Methods, GMM, Gaussian
Calorie-by-Photo (Python, PyTorch, ZoeDepth). 2026
Processes, Bayesian Optimization
⋄ End-to-end calorie estimation from a single photo, deployed as a
• Neural Networks, VAEs, Diffusion and
production Telegram bot. Pipeline: food instance segmentation
Generative Models
(Mask R-CNN, ResNet-50 backbone with FPN), food classifica-
• CNN, U-Net, Transformers tion (ConvNeXt-Tiny), depth-aided volume estimation via ZoeDepth,
• Model Fine-Tuning, CUDA, GPU mapped to caloric values. Trained and evaluated on public bench-
Training marks FoodSeg103 and FoodX-251.
• Linear Algebra, SVD/PCA, Fourier
Transforms GitHub
• PyTorch, TensorFlow, scikit-learn
• NumPy, Pandas, OpenCV
• statsmodels, Prophet, Matplotlib
• Python, Java, SQL, JS, R
• FastAPI, Java Spring, React, REST API
• PostgreSQL, MySQL, Redis
• Git, Docker