Ehsan Nazari

ML Researcher & CV Engineer | Generative Models · Transformer-Based Vision · Real-Time Perception

ehsan_green.jpg

At Pacefactory, I build and deploy real-time perception systems for industrial environments: object detection, keypoint estimation, and segmentation from raw camera feeds, trained, evaluated, and running in production.

I hold a Master’s in Applied AI from the University of Ottawa (GPA: 10/10, Vector Scholarship recipient), where my research focused on generative models and adversarial attacks on face verification systems under Professors Branco and Jourdan.

On my own time, I train generative models from scratch (diffusion and flow matching, DiT and VAE implementations, LoRA-finetuned multimodal LLMs) and study how deep networks learn, represent, and generate visual content. Recent project. I also write about object detection internals on Medium.