At Vyorius, I'm implementing computer vision improvements for trajectory estimation, reducing average drift from 3.1 km to 0.11 km through 4-corner geo-calibration. I also accelerated altitude optimization by about 12× and increased frame rate by more than 2× using pyramidal Lucas-Kanade flow and RANSAC affine estimation.
I've built production-style ML projects spanning a multi-LLM movie recommender, autonomous multi-organ medical image segmentation, and GAN-based speech bandwidth extension. My HybridGAN-BWE project achieved 38× real-time inference and a 67% parameter reduction, with MLflow tracking and an interactive Hugging Face Space for real-time audio enhancement.
I'm pursuing a B.Tech in Computer Science and Artificial Intelligence at IIIT Delhi and enjoy applying Python, PyTorch, LLMs, and computer vision to practical systems.
