Weiss Schwarz Simulator
A deterministic engine built for reinforcement learning: stable action encodings and fast batch stepping.
Denmark · Systems-first · Quietly experimental
I build deterministic sims and tools that feel sturdy under pressure.
My work lives at the intersection of rigorous engineering and play: card games, Pokémon battles, and reinforcement learning pipelines built to scale without losing clarity.
Rust hot loops
Python bindings
Deterministic RL envs
Selected work
Projects with sharp edges and clear intention.
RL Agent for Pokémon Showdown
PPO-based agent for Gen 9 Doubles using Stable-Baselines3 and poke-env, with offline imitation training and online evaluation flows.
More experiments
Smaller builds ranging from disk tooling to TypeScript utilities — always chasing clarity, speed, and just enough delight.
Process
How I keep complex work readable.
01
Model the rules
Build state machines that reveal intent early — deterministic, testable, and built for replay.
02
Expose clean surfaces
Action spaces, observation encodings, and APIs that stay stable while internals evolve.
03
Ship with documentation
Benchmarks, changelogs, and docs make the work feel solid for the next person.
Experiment bay
CV rendered as a training dashboard.
RUN_LOG :: vwp
Training playback
AP CS
Auditdata
Klimakampen
MSc + Gioia
Expected grad
Eval curves
Telemetry snapshots
Checkpoint vault
Certification
PADI Divemaster
Certified · diving discipline + calm under pressure.
Languages
Danish / English
Native Danish · fluent English.
Semester Abroad
Eindhoven, NL
Studied abroad with a focus on systems and collaboration.
Contact
Let’s build something with a sharp point of view.
Collaboration
I’m open to projects in simulation, tooling, or product engineering. If you want a collaborator who documents the sharp edges, I’m in.