At Alpha CLoud Labs, I co-developed real-time virtual advertising software that inserts geometrically stable ads into live PTZ and aerial camera feeds without external sensors or specialized tracking hardware.
I optimized the video processing pipeline to approximately 250 ms end-to-end latency for broadcast-quality sports applications. I also designed and implemented a custom YOLO inference engine in Rust, achieving approximately 40% higher throughput than comparable open-source Rust implementations.
I contributed to product validation with Formula 1, DTM, and FIS Ski using challenging live sports broadcast footage.
At Anvaya Innovations, I developed an image-processing framework that estimates fluoride concentration from smartphone images captured under UV illumination. I also built an OpenCV-based processing pipeline and evaluated classical computer vision and machine learning approaches.
