Projects

  • HSM-Signer Service (PDF/XML Signing and Verification with KamuSM)

    (56 ratings) 6428 views

    A secure and scalable HSM-backed digital signature infrastructure was developed for KamuSM-compliant XML (XAdES) and PDF (PAdES) signing/verification processes. With REST APIs, PKCS#11 integration, asynchronous queue architecture, and Docker-based deployment, the solution delivered high performance, regulatory compliance, and strong operational observability.

    Read more

  • Natro — Domain & Hosting Platform Modernization (10+ Services)

    (22 ratings) 3085 views

    More than 10 mission-critical VB.NET services, including domain activation, hosting provisioning, payment processing, customer management, and renewal workflows, were modernized and migrated to a more sustainable, observable, and high-performance .NET 7 architecture for one of Turkey’s high-traffic hosting platforms. This transformation reduced production error rates, improved API response times, shortened incident detection time, and significantly increased operational reliability in production environments.

    Read more

  • Octapull — Remote Video-Based Collaboration Platform

    (21 ratings) 3476 views

    Contributed to the development of a real-time video-based collaboration platform designed for remote interviews, field operations, digital channel management, and enterprise communication scenarios. With a Jitsi-based communication infrastructure, a high-concurrency service architecture, performance optimizations, and operational modules, the platform evolved into a reliable, scalable, and enterprise-ready solution.

    Read more

  • Parkinson’s Disease Detection Using Deep Learning

    (36 ratings) 4380 views

    A deep learning-based classification system was developed to detect Parkinson’s disease through voice signal analysis. In the project, multiple phonation recordings of the vowel “A” were analyzed to predict the presence of the disease. An end-to-end machine learning pipeline covering data preprocessing, feature extraction, model training, and evaluation was established, achieving approximately 92% classification accuracy.

    Read more