Кіріл

Kirill Kosak
Student of Systems Analysis | Aspiring Data Analyst / Python Developer
City: Brovary, Kyiv Oblast, Ukraine Phone: [open contact info](look above in the "contact info" section)
Email: [open contact info](look above in the "contact info" section) Telegram: @Wep0ut1337
LinkedIn: [open contact info](look above in the "contact info" section)
Project: app.cleanmetrics.pro
About Me
• First-year student of Systems Analysis at the State University of Trade and Economics (Kyiv), with a solid grounding in mathematical modeling and data processing.
• Finalist and semi-finalist of large-scale Ukrainian hackathons, plus an independently built and deployed SaaS product — demonstrating the ability to take a technical problem from concept to a working, security-conscious system.
• Actively work with AI-assisted development tools (Claude, Copilot) as part of a modern engineering workflow — while personally owning the architecture, debugging, and security decisions behind every project.
• Aiming to apply analytical thinking, SQL, and Python to data analytics and fraud-detection-adjacent roles.
Technical Skills
• Programming & Databases: Python (Pandas, NumPy, Asyncio, FastAPI), SQL, Microsoft SQL Server, PostgreSQL basics.
• Data Analysis & Visualization: Microsoft Excel (pivot tables, advanced formulas, data modeling), data cleaning, data visualization.
• Web & Backend: FastAPI (background tasks, rate limiting, webhooks), React, Docker, docker-compose, Nginx, REST API design.
• Security Awareness: IDOR checks, SQL-injection testing, CORS handling, anti-enumeration patterns, hashed one-time tokens, secrets hygiene.
• Tools: Visual Studio Code, Jupyter Notebook, Git, SSMS, PyQt6, Claude / AI-assisted development.
• Core Competencies: Systems analysis, sensor data fusion, fraud detection, mathematical modeling, asynchronous data processing, QA automation.
Hackathons & Competitions
Finalist | BEST HACKathon Lviv 2026
• Developed an Inertial Navigation System (INS) for GPS-denied environments.
• Implemented sensor fusion logic combining data from multiple IMUs (accelerometers and gyroscopes).
• Used an Error-State Kalman Filter (EKF) and barometric data to correct trajectory and minimize drift.
• Processed raw telemetry logs with Python and Pandas to generate 3D motion trajectories.
Semi-finalist | Monobank & Skelar Hackathon 2026
• Built a fraud-detection model analyzing large volumes of banking data (160K transactions, 40K users).
• Designed a 5-feature risk model (graph-based card-network analysis, payment-failure patterns, post-registration velocity, geo-mismatch, composite risk metrics) and a two-tier ensemble (130 bagged models + CatBoost), reaching F1 = 0.78.
• Proposed a business rollout strategy ("Scale&Trust") based on soft-decline instead of hard blocking, with human review for medium-risk cases — independently converging on the same approach later confirmed as industry practice in a real bank's fraud team.
Projects
CleanMetrics — Cookieless Web Analytics SaaS | Co-founder & Developer (Python, FastAPI, React)
A deployed, production-style SaaS product built end-to-end with a co-founder, covering the full stack — backend, database, frontend, billing, and security.
• Backend: Built domain-separated FastAPI services (auth, billing, sites, stats, tracking, Telegram bot) with background-task processing and a Gemini-based AI agent with tiered retry/fallback logic.
• Security: Ran a full codebase audit that found and closed three critical vulnerabilities; added 43 automated QA checks (IDOR, rate limits, CORS, SQL injection); implemented anti-enumeration password reset with SHA-256-hashed, single-use, time-limited tokens; replaced weak JWT/salt secrets with cryptographically strong ones; closed a .env leak into Docker images.
• Database: Designed and iteratively evolved the schema across 9 migrations; implemented cascading account deletion across all foreign-key-dependent tables inside a single transaction.
• Debugging: Diagnosed and fixed a silent production bug where tracking events were being dropped — traced to a CORS-preflight issue in `sendBeacon`, resolved by rewriting the client tracker to a fetch-first, preflight-free approach.
• DevOps: Containerized the full stack (Docker, docker-compose, Nginx), removed a legacy payment provider (Stripe) end-to-end in favor of a single regional processor, and authored launch/runbook documentation for deployment.
• Live at: app.cleanmetrics.pro
Automated Candidate Sourcing Tool | Developer (Python, Selenium/CDP)
Built a browser-automation tool to source job candidates from two major Ukrainian job boards, writing structured results to Google Sheets with live status tracking.
• Bot-detection workaround: Diagnosed that a login failure was not a cookie-encryption issue but active bot detection, and solved it by connecting via Chrome DevTools Protocol (CDP) to an already-authenticated browser session instead of automating a fresh login.
• Resilient parsing: Iterated through three approaches to reliably detect "viewed" / "contacted" candidate states on a site with unstable CSS classes, landing on a stable method based on unique SVG icon path fragments.
• Workflow tooling: Added a color-coded status dropdown column and a heuristic relevance filter (keyword + blacklist based) that flags candidates for review without deleting data.
Automated Data Processing & Deduplication System | Developer (Python/Asyncio)
• Built a high-performance tool for automated collection and cleaning of data from external APIs, integrating thousands of records while removing duplicates and enforcing format consistency without manual intervention.
• Performance: Implemented an asynchronous architecture allowing network requests without blocking the UI.
• Rate limiting: Designed a universal request-throttling module to keep API usage within service limits.
• Data quality: Built a name-deduplication algorithm accounting for transliteration, word-order variation, and typos, minimizing database noise.
• Architecture: Structured the system around Dependency Injection for easy scaling and new data-source integration.
Data Analysis & Optimization
• Performed cleaning and standardization of complex datasets using SQL and Excel.
• Built performance dashboards in Excel using Power Pivot to visualize business trends and insights.
Certificates

Semi-finalist, SKELAR x mono AI Competition (February–April 2026)

Certificate of Participation, BEST HACKathon Lviv 2026
Education
State University of Trade and Economics (SUTE) | 2025–2029
• Faculty of Information Technology
• Bachelor's in Systems Analysis
Languages
• Ukrainian — Native
• English — Intermediate (B1)