Zawar Qureshi
Senior Machine Learning Engineer at Niantic Spatial, Inc.
Stony Stratford, England, United Kingdom
Summary
👤
Senior🎓
Top SchoolZawar Qureshi is a Senior Machine Learning Engineer with 8 years of experience specializing in computer vision, generative AI, diffusion models, and 3D reconstruction. He has driven end-to-end research and production deployments at Niantic—authoring a CVPR 2025 paper on RGBD diffusion for 3D Gaussian splat stylization and delivering multi-fold speed and cost reductions across training and inference pipelines. Zawar combines deep research chops from Oxford VGG with pragmatic engineering: migrating production stacks to PyTorch Lightning and Ray to cut training time 3x and infrastructure costs by over 50%. He’s led teams to integrate on-device acceleration (Apple Neural Engine, Qualcomm) and advanced quantisation that halved latency and battery use for AR products. Comfortable across CUDA optimization, scalable training, and mobile deployment, he’s equally adept at debugging at scale—having fixed a critical bug affecting millions of user scans. Based in the UK, he blends academic rigor with product-focused systems work to bring cutting-edge vision models into real-world AR experiences.
8 years of coding experience
7 years of employment as a software developer
Master of Engineering (MEng) Engineering Science, Master of Engineering (MEng) Engineering Science at University of OxfordA-Levels Economics Mathematics Further Mathematics Chemistry Physics English (General), A-Levels Economics Mathematics Further Mathematics Chemistry Physics English (General) at Karachi Grammar School
English, Urdu, Arabic
Github Skills (55)
keras-neural-networks10
occlusion10
depth-estimation9
cubic9
densenet9
ray8
optimization8
arxiv8
python8
equation8
reinforcement-learning8
resnet8
parallel8
keras-models8
converter8
Programming languages (5)
Github contributions (5)
qureshizawar/CUDA-quartic-solver
Jun 2019 - Jun 2020
A general cubic equation solver and quartic equation minimisation solver written for CPU and Nvidia GPUs, for more details and results, see: https://arxiv.org/abs/1903.10041. The library is available for C++/CUDA as well as Python using Pybind11.
Contributions:3 releases, 18 commits, 19 PRs in 1 year
cudacpuequationpythonarxiv
qureshizawar/qureshizawar.github.io
Jun 2019 - Apr 2021
A lightweight, customizable single-page personal portfolio website template built with JavaScript and Sass
Contributions:35 PRs, 61 pushes, 36 branches in 1 year 10 months
sasspersonal-pagecssjavascriptcss3
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.