Skip to content
View ausstein's full-sized avatar
  • Constructor University
  • Bremen

Block or report ausstein

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ausstein/README.md

👋 Hi, I'm Yannick Holtkamp

I'm a physicist and computational scientist with a passion for integrating classical simulation algorithms with machine learning techniques to advance our understanding of complex physical systems.

During my PhD at Constructor University Bremen, I developed a fully differentiable implementation of quantum dynamics simulation methods (NISE) in PyTorch, combining them with neural networks to improve accuracy and scalability for biologically relevant systems. I also focus on optimizing scientific codebases for high-performance computing, enabling efficient exploration of complex parameter spaces.

I'm particularly interested in applying these methods to AI-based world models, combining differentiable physical models with modern AI architectures to build systems that not only learn from data but also leverage structured knowledge from natural laws.


🧩 Projects

Differentiable Quantum Dynamics Simulation

Built a PyTorch-based differentiable implementation of the Numerical Integration of the Schrödinger Equation (NISE), enabling integration of classical physics models with modern AI architectures.

🚗 Multiprocessing for Reinforcement Learning in Trackmania

Contributed to a GitHub project focused on reinforcement learning in the racing game Trackmania. Implemented a multiprocessing system that allowed multiple agents to run separate Trackmania environments simultaneously, significantly increasing training data collection speed.

While my changes were not merged into the main repository due to internal changes by the authors, they adopted a similar approach inspired by my implementation in their internal build.

🎮 Game Jam Projects

Participated in 14 game jams, building complete games in C# and Unity. Many are playable in browser!

Game jams have strengthened my rapid prototyping skills and creative problem-solving. Rather than focusing on top placement, I challenge myself with something new in every entry.

Highlights

🖥️ While loops <-- Browser Playable, click to play

Developed during Ludum Dare 47 - Theme "Stuck in a loop"

One of my favorites for tech audiences:

You are stuck in an infinite while loop, try to modify the variables to break out before the PC overheats.

💬 "This game feels like the closest I’ve ever come to actually programming something. Terrifying."

💬 "Really hits home for developers!"

💬 "Perfectly captures the anxiety of debugging loops."

💬 "reallyFinalyVersion7 is the best joke of the jam. Really hits home."

💬 "while (player.enjoyedGame()) { print(“Nice game!”); } Nice game! Nice game! Nice game! Nice game! … (Player died from overheating)"

While Loops received highly positive feedback at Ludum Dare 47 for its humor, creativity, and polished execution. Tech audiences especially appreciated the in-jokes and the clever use of while loops as a game mechanic.

While Loops Game

🦑 Deep Sea Mining <-- Browser Playable, click to play

Developed for Ludum Dare 48 — Theme: "Deeper and Deeper"

A submarine mining game focused on exploration, upgrades, and survival. Featuring a dynamic upgrade system, procedurally increasing difficulty, and post-jam improvements based on player feedback.

💬 "This is one of the few games I would sit down and play even if I wasn’t rating games for a game jam."

💬 "Great progression curve — never felt like I was grinding."

💬 "Very fun, deserves to be developed further!"

Players praised the polished feel, addictive progression, and atmospheric graphics. I continued to improve the game after the jam, incorporating player feedback to balance combat, improve lighting, and enhance controls

Developed during Ludum Dare 48 - Theme "Deeper and Deeper"

Go deeper and deeper, farm crystals, upgrade your submarine and defend yourself against hostile creatures.

Deep Sea Mining Image

🎮 Multiplosphere (Download for Windows)

Download link

Developed for Ludum Dare 54 — Theme: "Limited Space"

A fast-paced 3D arena shooter where multiplying spheres threaten to overwhelm the player. Focused on chaotic gameplay, fast decisions, and satisfying explosions. Collect upgrades, time your moves, and survive as long as possible!

💬 "Infinite replayability — would make a great time-waster on mobile!"

💬 "The spheres... they’re everywhere..."

💬 "Nice atmosphere, polished feel — well done!"

Multiplosphere Image

🧬 Personal Interests

  • Designing and maintaining complex, balanced ecosystems in aquariums and paludariums 🌿🐠
  • Exploring generative AI tools and AI agent design
  • Board game design — published "Sparta" at age 14 with Queen Games
  • Rowing, badminton, and running 🏃‍♂️

🔗 Connect with me

Publications

Main Author

Co-Author

Popular repositories Loading

  1. trackmania_rl_public trackmania_rl_public Public

    Forked from Linesight-RL/linesight

    AI Plays Trackmania with Reinforcement Learning

    Python 3

  2. MojoFreecellSolver MojoFreecellSolver Public

    1

  3. stable-diffusion_deforum stable-diffusion_deforum Public

    Forked from deforum/stable-diffusion

    Jupyter Notebook

  4. Ryven Ryven Public

    Forked from leon-thomm/Ryven

    Flow-based visual scripting for Python

    Python

  5. torchTT torchTT Public

    Forked from ion-g-ion/torchTT

    Tensor-Train decomposition in pytorch

    Python

  6. qutip qutip Public

    Forked from qutip/qutip

    QuTiP: Quantum Toolbox in Python

    Python

160 contributions in the last year

Contribution Graph
Day of Week April May June July August September October November December January February March
Sunday
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Less
No contributions.
Low contributions.
Medium-low contributions.
Medium-high contributions.
High contributions.
More

Activity overview

Contributed to CPBPG/TorchNISE, ausstein/MojoFreecellSolver, ausstein/ausstein and 5 other repositories
Loading A graph representing ausstein's contributions from April 07, 2024 to April 10, 2025. The contributions are 97% commits, 2% issues, 1% pull requests, 0% code review.

Contribution activity

April 2025

Created 7 commits in 1 repository
Created 1 repository
Loading