Kulbhushan Singh

Software engineer specializing in video quality and computer vision, with 3+ years building production C++ systems for broadcast, OTT, and surveillance pipelines at Interra Systems.

My work spans no-reference video quality assessment (VQA) for H.264, AV1, and MPEG2 streams; scene-change detection achieving 97% precision and 100% recall; and artifact detectors for blocky dropout, chroma dropout, and black bar detection with full 16-bit/4K support. I build optical flow pipelines using Lucas-Kanade, Farneback, and Horn-Schunck algorithms, fused with Kalman filters and RANSAC for robust motion estimation. Performance optimization via SIMD/Intel intrinsics has delivered 40–80% speedups across multiple modules.

I am an active open-source contributor to TorchMetrics: VIF metric reduction='none' support (PR #3226) and classification metric NaN fix (PR #3196).

Seeking: Visa-sponsored roles in Video Quality · Computer Vision · HPC Engineering — Europe (German speaker) preferred, open to UK, AU, CA, SG.