Paul-edouard Sarlin
Research Scientist at Google
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolPaul-edouard Sarlin is a research scientist at Google Geo with a decade of experience at the intersection of 3D computer vision, geometry, and deep learning, grounded in a PhD from ETH Zurich. He has a track record of moving research into practice through internships at Meta Reality Labs, Microsoft, Magic Leap and an earlier Google Visual Positioning Systems stint, and by contributing to influential open-source projects like COLMAP and HF-Net. His work blends algorithmic rigor—improving pose solvers and reprojection error models—with pragmatic engineering, packaging Python bindings and model export pipelines for reproducible deployments. Comfortable across backend systems, model evaluation, and visualization, he has also built large-scale processing pipelines for remote sensing and productionized ML components. Colleagues know him for combining careful geometric insight with hands-on software craftsmanship that bridges academic research and real-world spatial systems.
10 years of coding experience
Bachelor's degree Microengineering, Bachelor's degree Microengineering at EPFLPhD Student Learning & 3D Computer Vision, PhD Student Learning & 3D Computer Vision at ETH ZürichScientific Baccalauréat, Scientific Baccalauréat at Lycée Maintenon - HyèresExchange Year Electrical and Computer Engineering, Exchange Year Electrical and Computer Engineering at National University of Singapore
French, English, German
Github Skills (31)
pytorch10
c-language10
geometry10
data-pipelines10
python10
machine-learning-models10
command-line-interface10
machine-learning10
data-export10
exporter10
deep-learning10
tensorflow10
command-line10
computer-vision10
structure-from-motion10
Programming languages (8)
Github contributions (5)
ethz-asl/hfnet
Sep 2018 - Jul 2020
From Coarse to Fine: Robust Hierarchical Localization at Large Scale with HF-Net (https://arxiv.org/abs/1812.03506)
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:154 commits, 6 PRs, 10 pushes in 1 year 10 months
Contributions summary:Paul-edouard contributed significantly to the implementation and export of machine learning models, including SuperPoint, and other neural networks. They worked on creating scripts to export predictions. The user also improved the dataset interfaces and added visualizations related to the datasets. They evaluated models, assessed their performance, and improved the visualization of data.
arxivabshierarchicalscalefine
colmap/pycolmap
May 2021 - Aug 2022
Python bindings for COLMAP
Role in this project:Back-end Developer
Back-end DeveloperContributions:8 releases, 29 reviews, 48 commits in 1 year 3 months
Contributions summary:Paul-edouard primarily focused on extending the Python bindings for the COLMAP library. They added an interface to the GP3P algorithm, including refinement options. Further contributions include switching from angle to squared error in the generalized absolute pose estimation and integrating changes from upstream COLMAP. The user also created tools to build and package wheels for Linux and macOS.
structure-from-motionpythonpython-bindingsintrinsicsmotion-tracking
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.