Ahmad Jarrar Khan

I'm a master student at EPFL studying Computer Science. I'm interested in inferring the physical world from visual data. Currently, I'm working on creating explorable 3D worlds from images using Gaussian Splatting.

Previously, I was a research intern at Johnson Electric and Vyro.ai. I completed my undergraduate studies at NUST.

Email  /  CV  /  Scholar  /  Github  /  linkedin

Projects

TC3DGS compresses dynamic 3D Gaussian representations using pruning, mixed-precision quantization, and trajectory interpolation, enabling efficient real-time applications.

This work presents a real-time video stylization method using Neural Cellular Automata (NCA), achieving temporal consistency and efficiency. An interactive demo showcases its performance on browser on everyday devices.

This project investigates vision-based autonomous navigation in dynamic environments, enabling drones to navigate and evade obstacles using visual inputs. It introduces imitation learning and a novel Drone Dynamics Module (DDM), demonstrating robust and effective performance in challenging scenarios.

Internships

I designed a pipeline for calibrating QA systems using synthetic images generated by diffusion models. Additionally, I developed a chatbot powered by large language models (LLMs) with integrated document search and retrieval capabilities to ensure trustworthy question answering.

I developed solutions for image segmentation, depth estimation, and image super-resolution, tailored for image editing applications. Additionally, I optimized neural networks for various image processing tasks to enable efficient deployment on mobile devices.

Teaching

Built using Jon Barron's code.
Last updated: 19/12/2024