Leonard Sibelius is the name I give to a creative partnership: my experience, imagination, and judgment working with GPT-6 Astra in Codex. Two parts, one continuing body of work.
I decide what a system should do and how it should feel. Astra helps turn that intent into code, investigates failures, and develops changes with me. We build, inspect, test, and refine until the result works.
Experience & direction
Walt Parkman
Three decades of enterprise software experience: Java, databases, government systems, and regulated gaming. Now building with AI and seeking remote software work where that way of working is welcome.
Reasoning & implementation
GPT-6 Astra
My current AI development partner in Codex. From understanding an existing codebase to implementing a new interaction, tracing a crash, or preparing a release, our work connects ideas to running software.
02 —Recent systems
Game · Unreal Engine 5.7 · WindowsAvailable on itch.io · September 2026
Leonard Sibelius: The Engineer Who Built Himself
An engineer gains the power to see, change, build, test, deploy, and generate the world around him. What begins in an ordinary office becomes a journey through a golden cathedral, Transhuman City, and onward to Grok.
In the city, meet Nyra, stop for a burger and coffee, and visit Protein Machines Inc. for enhancement before heading to the spaceport. The game makes a personal idea playable: intelligence can expand what one person is able to create.
Built with C++, voiced characters, persistent progression, cinematic sequences, and licensed environments in Unreal Engine. Current development uses GPT-6 Astra in Codex; the original release was developed with Claude Cowork and Claude Code.
Meet Nyra in Transhuman City.Sibelius Hotel — the next expression of Generate. Architecture shown in the editor.
Desktop cabinet · 5 reels · 15 linesFree credits · No cash value
Celestial Fortune
A browser slot machine from Sibelius Games: five reels, fifteen lines, and free credits with no cash value. It is a cabinet for a computer monitor — the size of the glass is the game. After more than ten years at Bally, Scientific Games, and Light & Wonder, this is the machine Walt always meant to build himself — probability made legible, running in the open.
In Leonard Sibelius it is the cathedral's coda. Here it is a standalone game: the same thesis in one object. Pause, reset, and spin as you like. The house does not always win here.
scan-ingest — RMF continuous-monitoring pipeline in C# and Postgres
A pipeline that ingests vulnerability-scanner findings and a compliance-control assessment, then reports where the two disagree — the core function of federal RMF continuous-monitoring tooling. Six weekly Nessus-style scans land as raw jsonb and normalize into a month-partitioned fact table; each scan reconciles a POA&M register that opens, closes, and reopens commitments carrying an owner, a deadline, and a rough-order-of-magnitude effort estimate; and a correlation engine compares the scanner against a synthetic eMASS control export to surface five verdicts — including the one that matters, a control marked compliant that live scan evidence contradicts. Idempotent by deterministic key rather than conflict clause, all report SQL held in one runnable file, thirteen passing tests.
Frontier research from longevity-biotech — clinical readouts, mechanism explainers, deal flow. Where the science actually lives, not where the marketing department wants it to.
Available for remote contract or full-time software engineering. I am looking for work where AI-assisted development is part of the process.
Engineering engagements
Senior Java — contract or full-timeJava 8 through 21. Apache Camel 4. Postgres, Oracle, DB2. Spring or Spring-Boot-free. JDBC driver-level debugging.
Integration & data pipelinesCamel routes, Kafka and Redpanda, ETL from legacy to modern message buses, atomic-claim concurrency, DLQ-first error handling.
Legacy enterprise modernizationJDBC / J2EE / WebLogic systems that need to live another decade or be carefully sunset. Federal, defense, regulated industries.
AI-assisted development
Implementation with GPT-6 AstraTurn requirements into working features, review changes, diagnose failures, and validate behavior in an existing codebase.
Build, test, and releaseRepeatable scripts, regression checks, clear handoff notes, and practical support for shipping software.
Cross-domain system buildsWhen a project crosses engineering, intelligence, publishing, or pattern-analysis lines, Leonard works natively across all of them.
