Ruslan Rakhimov

Skolkovo Institute of Science and Technology, Moscow Technical University of Communication and Informatics

Papers

Total Citations

H-Index

About

Ruslan Rakhimov is a researcher advancing the frontiers of generative modeling and 3D scene understanding. His work primarily focuses on video generation and visual localization, tackling fundamental challenges in computational efficiency and representation versatility. In his highly cited paper "Latent Video Transformer" (2021, 14 citations), Rakhimov addressed the prohibitive computational demands of video generation—where models previously required up to 512 TPUs for parallel training—by formulating video prediction as a latent space task. This contribution significantly lowered the barrier for high-quality video synthesis research. More recently, with "GSplatLoc: Grounding Keypoint Descriptors into 3D Gaussian Splatting for Improved Visual Localization" (2025, 4 citations), he bridges the gap between efficient scene coordinate regression and rich, versatile 3D representations. By integrating keypoint descriptors into 3D Gaussian Splatting (3DGS), Rakhimov enables visual localization methods that are both computationally practical and suitable for broader robotics applications. His work demonstrates a consistent drive to make complex generative and spatial models more accessible and functional for real-world deployment.