Hi, I’m Sabarish Kuduwa.
I am a Technical Lead at Uber working on Document Understanding. I architect spatial models and Vision-Language Models (VLMs) to scale document transcription globally, operating beyond human baseline accuracy.
Previously, I was the Technical Lead at the robotics startup Ebots, where I built the core Perception Stack—spanning from 3D reconstruction to pose estimation for embodied systems.
My work sits at the intersection of multimodal reasoning and physical space. I am focused on mathematically forcing models to obey spatial constraints and dropping down into the C++/CUDA architectures that allow machines to perceive the physical world. This site is a clean-room repository for my technical deep-dives, architectural notes, and open-source experiments.
Core Interests
- Machine Learning & Deep Learning
- Computer Vision & 3D Reconstruction
- World Models & Embodied AI
- Robotics Perception
Feel free to reach out if you want to collaborate or discuss the math behind the metal