Markus Krimmel
Member Of Technical Staff at Perplexity
Berlin, Germany
Summary
🤩
Rockstar🎓
Top SchoolMarkus Krimmel is a software engineer and PhD student-turned-practitioner with a decade of experience building and hardening ML and RL infrastructure. Currently a Member of Technical Staff at Perplexity in Berlin, he blends research-grade rigor from his work at Max Planck institutes and ETH Zürich with hands-on engineering at startups like Jina AI. He has a strong open-source footprint in popular reinforcement learning projects (notably core contributions to OpenAI Gym and Gymnasium) where he improved core data structures, documentation, and test suites that many RL researchers rely on. Markus specializes in vectorized environments and compatibility work for PyTorch-based RL libraries such as Tianshou, reflecting a pragmatic focus on scalability and reproducibility. Colleagues know him for turning academic ideas—geometric generative models and symbolic skill representations—into robust, test-covered code. He brings a rare mix of mathematical depth and production-oriented software craftsmanship to ML systems.
9 years of coding experience
Master of Science - MS Mathematik, Master of Science - MS Mathematik at ETH Zürich
Github Skills (19)
environmental10
pytorch10
dev-environment10
pytest10
python10
testing10
gymnasium10
openai-gym10
reinforcement-learning10
enviroment10
numpy10
api10
environ10
unit-testing9
documentation9
Programming languages (5)
Github contributions (5)
openai/gym
Jan 2022 - Aug 2022
A toolkit for developing and comparing reinforcement learning algorithms.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:68 reviews, 18 commits, 18 PRs in 7 months
Contributions summary:Markus primarily contributed to the core functionality of the `openai/gym` repository, focusing on enhancements and documentation improvements. Their work included implementing simplified representations for the Box space, a fundamental data structure within the library. The user also addressed minor issues such as suppressing PyGame import messages and adding kwarg documentation for certain environments. Furthermore, they made extensive changes to docstrings across several MuJoCo environments, improving clarity and providing detailed information about the environments.
comparingreinforcement-learning-algorithmsdevelopingdeep-learningreinforcement-learning
thu-ml/tianshou
Nov 2021 - Dec 2022
An elegant PyTorch deep reinforcement learning library.
Role in this project:ML Engineer
ML EngineerContributions:3 reviews, 6 commits, 8 PRs in 1 year 1 month
Contributions summary:Markus made significant contributions to the `tianshou` library, focusing on enhancements related to vector environments. They implemented and refined functionalities like `set_env_attr` and `get_env_attr` for vectorized environments, crucial for efficient reinforcement learning. Furthermore, the user adapted the library to support the new Gymnasium API and incorporated necessary updates to maintain compatibility with the latest versions of Gym. They also addressed and resolved various issues to improve the library's overall functionality, testing and codebase.
deep-reinforcement-learningbenchmarknpgtd3reinforcement
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.