Harsha Kalinga

On DefenXion, I developed an AI-powered cybersecurity platform for near real-time threat detection and automated response. I implemented Random Forest and Gradient Boosting models using the CICIDS2017 and UNSW-NB15 datasets, achieving 99.88% detection accuracy with a 0.125% false positive rate.

I built the platform with FastAPI, MongoDB, React, and React Native, including JWT authentication, role-based access control, real-time alerts, and analytics dashboards.

On FurniVision, I developed 3D room-planning features with React Three Fiber and Three.js, including interactive furniture placement. I also designed and implemented 11 frontend pages and contributed to its Node.js, Express, and MongoDB backend.

At Innodata Lanka, I perform AI data annotation and evaluation, assessing generated audio and text for quality, accuracy, naturalness, and consistency. I hold a BSc (Hons) in Computer Science from the University of Plymouth and am interested in growing my expertise in backend engineering.