Skip to content
P

Projects

Work I can show in public. Most of my employed work is internal, so it lives on the index as outcomes instead.

Secret Hitler XL

Sep 2026 – ongoing · Live

An online board-game platform for 5 to 19 players with ML bot players, built and run solo. I owned product, architecture, plans and reviews; AI coding agents implemented under that direction.

decisions in the latest self-play set
38.2M
decisions in the latest self-play set
model generations trained and gated
8
model generations trained and gated
bot calls replayed identically, Python vs Rust
1.84M
bot calls replayed identically, Python vs Rust
test cases across the main suites
~9,600
test cases across the main suites
Platform
  • Event-sourced, deterministic Python engine: live games resume after a restart, replays are exact, and each player receives only what they may know.
  • FastAPI and WebSockets with a long-poll fallback, idempotent retries and an offline queue, on SQLite (WAL) with versioned migrations.
  • Installable PWA from 320 px to desktop; OAuth with only an HMAC of the provider id stored, self-service GDPR erasure, moderation and an audit log.
ML bot players
  • A heuristic teacher at three difficulty levels generates self-play data: 1.01M games, 38.2M decisions in the latest set.
  • Eight model generations so far; the latest plays every difficulty level, in the base game and XL.
  • Transformer policy and value networks trained by behaviour cloning, from ~1M to ~4.7M parameters, tracked in MLflow.
  • Each generation is promoted only through a pre-registered statistical gate: A/A baseline, bootstrap confidence intervals, margins fixed before any candidate game.
  • Exported to ONNX (parity checked to 5e-4) and served from a separate bot service with leases, heartbeats and capacity limits.
  • A live bot decision takes 15 ms median and 82 ms p99 on one CPU thread; a load test with 72 bot seats at 4x pace held p99 at 140 ms with no late moves.
  • PPO self-play on top of the cloned policy, anchored to it with a KL penalty and trained against a league of past snapshots, one difficulty level at a time.
  • Switching Kaggle T4 training from emulated bf16 to native fp16 made it 4.5x faster (231 to 1,029 rows/s) and cut peak GPU memory from 7.0 to 1.5 GB.
Table talk
  • Bots talk at the table through an LLM with provider fallbacks, or a small local model fine-tuned for the job.
  • Distillation: a 27B open-weights teacher wrote lines for 8,898 statements, filtered by strict screens and a fact judge. A 2B student was fine-tuned on them with LoRA (rank 16) in 16 minutes.
  • Merged and quantised from 4.6 GB to a 1.2 GB GGUF, it runs on llama.cpp at a 259 ms median with 81.8 % of lines judged correct.
Rust port, in progress
  • Game engine 65 to 80 times faster than Python on the engine kernel.
  • Heuristic bot ported with byte-exact parity: 1.84M recorded calls replay identically and 60+ planted mutants are each caught by a named check.
  • Native data generation merged: the Rust loop reproduces the Python training data byte for byte (1.28M rows, 603 MB). The encoder runs 3.35x faster and the legal-move mask 12.4x.
  • A typed bot port was stopped by its own gate: it measured 1.22x where the pre-registered rule needed about 10x, and two reviewers confirmed the no-go.
Operations
  • Self-hosted on my homelab; the public site runs on a Raspberry Pi 5.
  • Multi-arch Docker images in a self-hosted registry. Every merge passes a public-export leak gate.
  • Two independent model reviewers check every plan and diff; fixes come with tests that fail first.

Homelab & this site

2021 – ongoing · Running

A personal server practice since 2021: VMs, containers, monitoring and alerts, a media stack, and SSO for an app behind a proxy. This site is built and deployed on it.

LinuxDockernginxVMwareWireGuard

urek.eu

RimWorld mods, agent-built

Oct 2026 · Proof of concept

Three open-source proofs of concept for RimWorld 1.7, written by AI coding agents under my direction and reviewed by Claude and GPT models. Tested in a headless in-game harness; not line-by-line human reviewed and not released on the Workshop.

behaviour-exact optimizations in ColonyScale
38
behaviour-exact optimizations in ColonyScale
What the three mods do
  • ColonyScale: performance patches that keep vanilla's decisions and random-number use identical. Each patch fingerprints the code it targets and switches itself off after a game update, or when another mod patches the same code.
  • LivingWorld: a Dwarf Fortress-style simulation where settlements and factions trade, fight and change on their own.
  • RimPrison: a Prison Architect-style game mode with intake, regimes, staff and reports.

Manufacturing app: auth and admin

Jul – Aug 2025 · Private

Rebuilt authentication and administration for a private manufacturing (MRP) web app, so shop-floor workers can sign in fast and admins can manage people and permissions themselves.

sign-in methods: password, PIN, NFC card
3
sign-in methods: password, PIN, NFC card
permission tiers
6
permission tiers
What changed
  • Moved the database from SQLite to MySQL running in Docker.
  • Password, PIN and NFC-card login, so workers on the floor do not type passwords.
  • Six-tier role-based access with an admin dashboard for users and permissions.
  • Internationalisation and a dark mode.

MySQLDockerRBACNFCi18n

Steam Workshop Monitor

2022 · Archived

An open-source Discord bot that told server operators when their Workshop mods updated, built to solve my own server's problem. Reached ~100 servers, the pre-verification ceiling. Early code, kept as an origin story.

servers
~100
servers
GitHub stars
24
GitHub stars

Telegram-To-Discord

2022 · Archived

Mirrors Telegram channels into Discord through a webhook, forwards photos and short videos, and translates anything not in English.

GitHub stars
51
GitHub stars
P2

Upstream contributions

>esc

Keyboard

j k
Next / previous line
o
Open or close the line
Ctrl K
Command palette
t
Toggle dark mode
?
This help