Full-stack software engineer based in Tirana, Albania.
I build reliable web applications and backend services, working across frontend interfaces, APIs, databases, system integrations, automated testing, and delivery pipelines. My experience includes enterprise application development, real-time workflows, and the modernization of existing systems.
I value clear architecture, maintainable code, security, and practical solutions to complex business problems. I am comfortable learning new technologies when they are the right fit for the system.
Languages: C#, TypeScript, JavaScript, Python, SQL
Backend and integration: ASP.NET Core, .NET MVC, Entity Framework Core, REST APIs, WebSockets, SignalR, JWT/RBAC, Keycloak, microservices and event-driven systems
Web applications: React, Vite, responsive interfaces, accessibility and progressive web applications
Data and applied ML: SQL Server, PostgreSQL, Supabase, scikit-learn, PyTorch, Transformers and model evaluation
Quality and delivery: Docker, GitHub Actions, Azure DevOps, Playwright, Vitest, xUnit/NUnit, automated testing and CI/CD
A secure employee, project and task-management platform demonstrating production-oriented API design and authorization boundaries.
- Uses JWT authentication, role-based authorization and current-user revalidation.
- Enforces project membership, task-assignment and administrative business rules.
- Includes password migration, login rate limiting, strict CORS, pagination, health checks and Problem Details responses.
- Uses explicit Entity Framework Core migrations and isolated integration-test databases.
- Provides 90+ xUnit integration-test declarations.
- Runs locked restore, formatting, Release builds, tests, migration checks and dependency auditing in CI.
Engineering themes: API security, authorization, data integrity, testing, migrations and maintainable backend architecture.
A production-minded, four-class sentiment-analysis system designed around trustworthy evaluation rather than inflated random-split results.
- Evaluates 71,117 records across 32 target entities.
- Prevents tweet-ID, exact-text and normalized-text leakage between evaluation folds.
- Achieves 0.6355 macro-F1 with Twitter-RoBERTa across five canonical grouped folds.
- Improves macro-F1 by 11.76% over the strongest classical baseline.
- Includes nested evaluation, class-level error analysis, model and data cards, CLIs and a browser demo.
- Provides 30 tests, 90%+ coverage, Python 3.11-3.14 CI, dependency auditing and a hardened container workflow.
Engineering themes: machine learning evaluation, reproducibility, data leakage, NLP, testing, CI/CD and deployment tradeoffs.