Lars Mescheder
Staff Research Scientist at Apple
Tübingen, Baden-Württemberg, Germany
Summary
🤩
Rockstar🎓
Top SchoolLars Mescheder is a Staff Research Scientist specializing in computer vision and 3D perception, with 11 years of experience building robust representation learning and reconstruction systems. He has moved research into production at scale—most recently developing core perception algorithms for Amazon Scout and now leading research efforts at Apple in Tübingen. His PhD from the University of Tübingen and a strong math/CS background underpin contributions to influential open-source work like Occupancy Networks, where he implemented TSDF fusion, data pipelines, and numerical robustness fixes for float16. Lars combines deep theoretical knowledge with hands-on engineering that spans sensors, fusion algorithms, and deployment on robotic platforms, and he often anticipates practical failure modes (e.g., adding noise to avoid float16 numerics) that keep systems reliable in the real world.
11 years of coding experience
9 years of employment as a software developer
Dr. rer. nat. (PhD) Computer Science, Dr. rer. nat. (PhD) Computer Science at University of TübingenStudent exchange Computer Science, Student exchange Computer Science at EPFLMaster of Science - MS Mathematics and Computer Science, Master of Science - MS Mathematics and Computer Science at Technische Universität Braunschweig
English, French, German
Github Skills (8)
computer-vision10
machine-learning10
python10
numpy10
3d-reconstruction10
data-preprocessing9
data-loading9
pytorch8
Programming languages (7)
Github contributions (5)
autonomousvision/occupancy_networks
Apr 2019 - Nov 2019
This repository contains the code for the paper "Occupancy Networks - Learning 3D Reconstruction in Function Space"
Role in this project:
ML Engineer
ML EngineerContributions:19 commits, 17 pushes, 1 branch in 7 months
Contributions summary:Lars contributed to the project by implementing and modifying core components related to 3D reconstruction. They developed a TSDF fusion algorithm, crucial for reconstructing 3D shapes from depth maps. Further contributions included adding data loading capabilities and correcting a misnomer in evaluation scripts. Finally, the user modified the point field to include a noise addition step to prevent numerical errors in float16 values, along with the addition of missing files related to data handling.
structure-from-motionreconstructionoccupancycomputer-vision3d-reconstruction
LMescheder/AdversarialVariationalBayes
May 2017 - Apr 2018
Contributions:61 commits, 8 pushes, 1 branch in 11 months
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.