Software Engineer specialized in AI systems, building self-hosted infrastructure, scalable backend/frontend applications, and production-ready machine learning pipelines. I focus on creating systems that operate independently, avoid vendor lock-in, and maintain full data control.
My expertise covers the full lifecycle: from GPU-optimized model deployment and high-performance APIs to CI/CD pipelines, monitoring, and automated testing. I design systems with reliability, scalability, and quality at the core, integrating strong QA practices across every layer.
Projects like Kurai-Transcribe, KuraiMusik, and KurAI2Video reflect this approach—AI systems running 24/7 with zero API costs, optimized performance, and production-grade stability, combining advanced AI capabilities with solid engineering and automation.
Production-grade self-hosted transcription API processing millions of minutes monthly. Custom vocabulary injection, named entity recognition with SpaCy, automatic punctuation restoration. Advanced GPU memory management with semaphore-based VRAM allocation prevents OOM errors. Voice activity detection + Pyannote speaker diarization. Real-time monitoring dashboard with Prometheus metrics.
Autonomous content generation system running 24/7 with zero human intervention. Dynamic content buffer system powered by advanced generative AI. Circuit breaker patterns ensure fault tolerance and reliability. Intelligent content orchestration with real-time metadata management and automated scheduling.
Enterprise-grade self-hosted video generation platform using state-of-the-art Wan 2.1 T2V-1.3B model. Eliminates cloud API costs that charge $0.15-0.40 per second—10,000 videos would cost $15,000-40,000. Advanced GPU memory management with dynamic CPU↔GPU offloading prevents OOM errors. Strict concurrency control ensures one active job at a time for predictable performance. Built-in monitoring dashboard tracks GPU utilization, VRAM usage, and model performance in real-time. Production-ready Docker deployment with automated health checks and failover mechanisms.
100% private personal diary app for Android. No accounts, no subscriptions, no trackers. Your entries never leave your device. Minimalist interface, emotional analysis, dark mode, and works offline. Developed with privacy as the top priority.
Available for AI infrastructure consulting and development projects.
Specializing in self-hosted AI systems, MLOps platforms, and production ML infrastructure.
Let's build AI systems that actually work at scale.