Darius Singh

Machine Learning Engineer focused on building and scaling production ML systems, with a strong emphasis on computer vision, scalable systems, and real-time inference. At i2V Systems, I lead development of large-scale computer vision infrastructure for real-time surveillance systems. I build end-to-end automated computer vision pipelines covering model development, training, evaluation, and deployment. This includes distributed face recognition systems trained and evaluated on 250M+ samples, and real-time edge-optimized inference systems running across 100+ edge nodes, processing 2000+ live surveillance video streams. I have built and scaled distributed training, inference, and evaluation systems that reduced runtime and compute by 30% , while improving face recognition performance from 94.5 to 97.41 TAR at FAR 1e-5 on IJB-C benchmarks, along with a 23% improvement in unconstrained end-to-end surveillance face recognition. Previously, I co-founded an AI startup where I built and deployed machine learning systems across e-commerce, advertising, and media, including recommendation systems, virtual try-on, real-time video analytics, and emotion recognition pipelines. My interests lie in applied machine learning, system design, and scalable ML infrastructure. I am currently working with self-supervised representation learning and multimodal models to improve robustness and generalization in real-world, high-variance environments.