Invited Talk: Department of Mathematics
Posted: Sun Oct 12, 2025 6:40 pm
Dear All,
An invited talk by Dr. Anurag Ranjan, Chief Scientist (Deepnight, USA), is scheduled as follows:
Date: 13-10-2025 (Monday)
Time: 6.00 PM
Venue: 6164
Title: "What I Cannot Create, I Do Not Understand: An Intuitive Take on Generative AI"
Abstract: Generative AI has exploded into the mainstream, powering billion-dollar companies and reshaping industries from art and design to drug discovery and robotics. But how do these models actually work? In this talk, we’ll take an intuitive journey through the fascinating world of generative models. We’ll start with the big picture, exploring why these models matter and how they create value, then peek under the hood to examine core ideas like diffusion models and flow-based models. Along the way, we’ll build an intuitive mathematical and logical understanding of how these models learn, focusing on representation learning and how they model extremely high-dimensional data to enable sampling from a regularized distribution. The goal is to connect these abstract concepts to their real-world implementations, developing a clear mental model of how generative systems are trained, scaled, and deployed to power today’s AI breakthroughs.
About the Speaker: Dr. Anurag Ranjan is the Chief Scientist at Deepnight, USA. At Deepnight, he is leading the development of the next generation of night-vision technologies. His work sits at the intersection of computer vision, generative AI, and hardware–software co-design, pushing the boundaries of what is possible in low-light and real-time imaging systems. Previously, Dr. Anurag was a senior research scientist at Apple, where he built foundational technologies deployed across millions of devices. He was the lead creator of MobileOne, the fastest convolutional neural network for mobile deployment, and FastViT, a state-of-the-art vision transformer optimized for speed and efficiency across Apple’s ecosystem. He also contributed to advances in neural rendering, powering key elements of personas in the Apple Vision Pro. Anurag earned his PhD at the Max Planck Institute for Intelligent Systems, where his thesis focused on the geometric understanding of motion. His research pioneered some of the earliest deep learning architectures for optical flow and structure-from-motion, establishing a foundation for modern computer vision.
All of you are invited to attend the talk.
Thank you.
Head, Department of Mathematics
An invited talk by Dr. Anurag Ranjan, Chief Scientist (Deepnight, USA), is scheduled as follows:
Date: 13-10-2025 (Monday)
Time: 6.00 PM
Venue: 6164
Title: "What I Cannot Create, I Do Not Understand: An Intuitive Take on Generative AI"
Abstract: Generative AI has exploded into the mainstream, powering billion-dollar companies and reshaping industries from art and design to drug discovery and robotics. But how do these models actually work? In this talk, we’ll take an intuitive journey through the fascinating world of generative models. We’ll start with the big picture, exploring why these models matter and how they create value, then peek under the hood to examine core ideas like diffusion models and flow-based models. Along the way, we’ll build an intuitive mathematical and logical understanding of how these models learn, focusing on representation learning and how they model extremely high-dimensional data to enable sampling from a regularized distribution. The goal is to connect these abstract concepts to their real-world implementations, developing a clear mental model of how generative systems are trained, scaled, and deployed to power today’s AI breakthroughs.
About the Speaker: Dr. Anurag Ranjan is the Chief Scientist at Deepnight, USA. At Deepnight, he is leading the development of the next generation of night-vision technologies. His work sits at the intersection of computer vision, generative AI, and hardware–software co-design, pushing the boundaries of what is possible in low-light and real-time imaging systems. Previously, Dr. Anurag was a senior research scientist at Apple, where he built foundational technologies deployed across millions of devices. He was the lead creator of MobileOne, the fastest convolutional neural network for mobile deployment, and FastViT, a state-of-the-art vision transformer optimized for speed and efficiency across Apple’s ecosystem. He also contributed to advances in neural rendering, powering key elements of personas in the Apple Vision Pro. Anurag earned his PhD at the Max Planck Institute for Intelligent Systems, where his thesis focused on the geometric understanding of motion. His research pioneered some of the earliest deep learning architectures for optical flow and structure-from-motion, establishing a foundation for modern computer vision.
All of you are invited to attend the talk.
Thank you.
Head, Department of Mathematics