Javad AMIRIAN

* Developing Volumetric Capture Algorithms * Building Digital Twins with Gaussian Splattings and Neural Rendering * Developping Computer vision pipelines for multi-camera capture systems, including synchronized acquisition, calibration, data preparation, and quality control. * 3D reconstruction and neural rendering workfows for digital humans and XR, with emphasis on robustness.

Research on HRI & Social Robotics | Member of euROBIN project Focusing on Computer Vision & Machine Learning models for Robot Navigation Bringing Large Visual & Language Models (LLM & VLMs) to Human-Robot Interaction

Vive Tennis is an ambitious project to improve the tennis experience with smart robots. As the head of AI, I have been responsible to build up an agile R&D team, plan the development and deliver the MVP product, design and implement the computer vision and navigation algorithms of Vive Tennis. * Built real-time perception stack for object detection, stereo/depth estimation, tracking, localization/SLAM, and navigation under latency and robustness constraints. * Deployed and optimized deep learning models on embedded NVIDIA Jetson using TensorRT/CUDA, focusing on inference speed, memory usage, reliability, and hardware cost constraints.

Visual Recommendation Engine for Furniture Retailers: * Developed visual AI prototypes for furniture understanding and recommendation, covering data preparation, model training, evaluation, and cloud-served ML workows. * Worked with deep learning models for visual retrieval/recommendation and related computer vision tasks such as detection, segmentation, and classifcation.

PixBall was an AI startup providing smart solutions for sports video analysis. My task was to develop image processing and machine vision algorithms and provide high-level API for the UI engineers. I contributed to PixArt, a video processing software for embedding graphical overlays and virtual advertisements into sports content. * Developed multi-camera calibration, lens distortion correction, 3D reconstruction, and real-time tracking components for production sports-broadcasting systems. * Implemented geometry-heavy computer vision modules in C++/OpenCV and Matlab under accuracy-sensitive and real-time constraints.