Mykyta

Mykyta Sapronov
Data Scientist / CV Engineer
Zaporizhzhia, Ukraine | [відкрити контакти](див. вище в блоці «контактна інформація») | [відкрити контакти](див. вище в блоці «контактна інформація») | github.com/G10BAL | Telegram:
@Mykyta_Sapronov | LinkedIn: Mykyta Sapronov

Summary
Computer Science student specializing in Algorithmic CS, focused on the mathematical foundations of Machine Learning
and Computer Vision: the calculus, linear algebra, and statistics behind the algorithms, applied to reproducible data
workflows.

Education
Wrocław University of Science and Technology Wrocław, Poland
B.Sc. in Algorithmic Computer Science 10/2024 – 02/2028
• Coursework: Statistics & Machine Learning, Algorithms & Data Structures, Cryptography, Abstract Algebra
Gymnasium No. 28 Zaporizhzhia, Ukraine
Secondary Education 09/2013 – 05/2024
• 3rd place globally – International Mental Arithmetic Olympiad; 3rd degree diploma – All-Ukrainian IT Olympiad
(final stage)

Experience
Self-employed Remote
Maths Tutor for Secondary Education Students 09/2025 – Present
• Prepared students for the Ukrainian NMT exam, raising a student’s timed practice-test score from 3/32 to 19/32
correct answers
NDA Recruitment – CV Summer School Theoretical track
Participant 07/2026
• Theory: camera models, classical & deep-learning keypoint extraction, normalization/embedding models, object
detection & classification, motion analysis & tracking, Vision Transformers

Projects
UAV Image Matching R&D – NDA Recruitment EPS Hackathon
Python, OpenCV, SuperPoint, LightGlue, LoFTR, RoMa, MAGSAC++ 09/28/2026 – 10/03/2026
• Investigated sparse, detector-free, and dense image matching methods for UAV reference-image localization under
viewpoint and appearance changes
• Compared matching latency and geometric consistency across models, identifying a trade-off between fast
SuperPoint+LightGlue matching (∼60 ms) and stronger but slower RoMa results (∼810 ms) on an exploratory
benchmark
• Analyzed unsuccessful localization cases using MAGSAC inlier counts and geometric acceptance criteria, showing that
increasing keypoint density alone did not resolve cross-view matching failures
Linear Regression from Scratch | Python, NumPy, scikit-learn 05/2026 – 06/2026
• Implemented linear regression from scratch: vectorized preprocessing, custom Gradient Descent and Normal Equation
solvers, OOP design
Wine Quality Analysis | Python, Pandas, scikit-learn, Statsmodels, Seaborn
• Performed EDA on the UCI Red Wine Quality dataset and built/compared classification models (Random Forest,
Logistic Regression, KNN, OLS); best binary model reached 91.9% accuracy, 0.76 macro-F1
Weapon Inventory Database Design | MariaDB, SQL
• Designed a normalized (3NF/BCNF) relational schema for an inventory system with triggers and role-based access
control

Skills
Programming: Python (OOP, design patterns, NumPy, scikit-learn), SQL
Computer Vision: OpenCV, SIFT, SuperPoint+LightGlue, LoFTR, RoMa, MAGSAC++, keypoint matching, image
normalization
Data & ML: Pandas, Statsmodels, Excel, Power BI, Machine Learning fundamentals, Statistics, data pipelines / ETL
concepts
Databases: MariaDB (schema design), MongoDB (NoSQL concepts)
Cloud & Infra: AWS (S3, EC2, IAM fundamentals), Git, Raspberry Pi (home-lab), ESP8266/ESP32
Languages: Ukrainian (Native), English (B2), Polish (B2)