Masaya Tsuji

Architected LLM-powered FastAPI services and RAG pipelines that cut manual ticket handling by 30% and reduced time-to-answer from 1.2 to 0.6 minutes while lowering P95 latency by 40% and error rates by 25%. Integrated AI endpoints into C#/.NET services, reducing integration defects by 35% and enabling phased rollouts for 20+ client tenants.

Optimized ML inference by porting Python to C++ and using ONNX Runtime with TensorRT to boost throughput to 30 FPS and improve performance by ~50%, and designed a counting algorithm that raised accuracy by 20%. Introduced unit testing and validation into the ML pipeline to improve stability and maintainability.

Refactored React/Redux into reusable hooks reducing bundle size 25% and improved LCP to 1.9s; built Flask REST APIs and resilient frontend integrations that reduced client-visible failures by 30% and improved data freshness SLAs to <2 minutes, increasing weekly active users by 22%.

Delivered React patient portals for clinics that increased self-service usage by 35% and reduced call-center volume by 25%; launched e-commerce storefronts that improved conversion by 18% and mobile checkout speed by 30% while creating a 30+ component UI library to cut build time by 40%.

Pursuing a Bachelor of Multimedia at the University of Ljubljana within the Faculty of Computer and Information Science / Faculty of Electrical Engineering, coursework focused on multimedia and computing fundamentals.