Engineered production EdTech applications including a K-12 homework management platform and a bilingual (English/Spanish) science & math learning game, deployed to real students at underserved schools via non-profit partners. Architected modular backend services (presentation, domain, data layers) in an Agile development environment, with 85%+ unit test coverage enforced through rigorous code review standards. Automated CI/CD pipelines using GitHub Actions across multi-platform builds, cutting deployment time by ~40% and enabling reliable, rapid release cycles. Optimized API performance and database interactions, sustaining sub-200ms response times across 1K+ concurrent user sessions under real production load. Led 30+ peer code reviews and resolved 50+ pre-release issues, reducing post-deployment bugs by ~25% and authoring technical documentation to improve team-wide knowledge sharing.

Built a real-time UAV collision-prevention system using YOLOv3 in C++, achieving 92%+ mAP at 30 FPS; prototyped stereo-vision pipelines processing 60+ FPS for navigation at 20m/s. Improved power line detection precision by ~35% by combining deep learning with classical vision techniques including Gabor filters, PCNF, and Hough Transform. Explored multi-sensor fusion (vision + IMU + LiDAR), reducing localization drift by ~20% and improving environmental awareness in dynamic flight conditions.