Soroush Saryazdi

Papers

Total Citations

H-Index

About

Soroush Saryazdi is a robotics researcher pushing the boundaries of how machines perceive and map the world. His work centers on two key areas: open-set multimodal 3D mapping and differentiable SLAM (Simultaneous Localization and Mapping). In his highly innovative paper "ConceptFusion: Open-set Multimodal 3D Mapping" (2023, 4 citations), Saryazdi tackles a critical limitation of traditional mapping systems—their inability to reason beyond a predefined set of concepts. By integrating open-set semantic understanding with 3D maps, he enables robots to interact with and navigate environments using unconstrained, natural language concepts, a leap forward for autonomous systems operating in dynamic, real-world settings. Earlier, with "gradSLAM: Automagically differentiable SLAM" (2019, 3 citations), he addressed the challenge of blending deep representation learning with the modular, complex nature of SLAM systems. By making SLAM differentiable, Saryazdi opened the door for end-to-end learning pipelines that can optimize mapping and localization from raw sensor data. Though early in his career, his contributions are already shaping the future of intelligent robotics, offering tools that make robots more adaptable, context-aware, and capable of understanding the world as humans do.