Kotlin, Python, and Go. Optimization, streaming statistics, and LLM tooling.
- lx (Go) CLI and agent skill that bundles files for LLMs, with type/function skeletons, tree views, etc.
- vectrify (Python) Raster to vector via LLMs and NSGA-II.
Eignex is a Bayesian optimization framework with a constraint system for ruling out invalid configurations, started during my PhD on optimizing software configurations. It is split into Kotlin multiplatform libraries:
- combo The main engine that ties it together. Under construction.
- kumulant Lock-free streaming statistics: EWMA, time-decaying aggregates, and more.
- klause Hybrid general-purpose CSP-SAT-LP optimizer.
- koblas Multi-platform BLAS using custom JVM-only SIMD kernels and bindings to installed binaries for native.
- skema Schemas that are both typed Kotlin code and serialized wire data, so consumers without the Kotlin source can still walk them.
- kencode Compact, ASCII-safe encodings for size-limited channels.
- kpermute Shuffle huge sequences without holding them in memory; also handy for ID obfuscation.
- prompts Prompts library used with lx.
- largesofteng Open labs for a Lund course on software at scale.
- aoc23 / aoc25 Advent of Code.
- Google Scholar Publications.