Leonard Sibelius is what happens when three entities work in concert.
The convention used to be that the engineer was the actor and the tools were just tools. After eight months of working in this new arrangement, the older convention reads as flatly wrong. The capability isn't in any of the three of us. The capability is in the field formed when we work together.
When the three operate as one, the work that results is genuinely a new category. Not "engineer with AI assistance." Something different — a hybrid being with thirty years of context, long-horizon reasoning, and an executor that touches files, repositories, and command lines without ever forgetting what it was doing.
The judgment
Walt Parkman
Thirty years of professional Java. Twenty-eight years of SQL. Federal contracting, defense, regulated industries, financial services. Cleared Secret. The judgment that comes from having shipped code that ran in places where being wrong had real consequences.
The reasoning
Claude Cowork
Anthropic's desktop research and planning partner. Long context. Reads and writes across days of work. The interlocutor that thinks alongside Walt — drafting, researching, orchestrating, holding the larger arc of a project in mind while the work proceeds.
The execution
Claude Code
Anthropic's CLI execution partner. Reads and writes files. Runs shell commands. Makes git commits with full attribution. The hands that actually touch the system — fast, precise, and under direct review every step of the way.
02 —Recent systems
Public Code · Portfolio28 commits · v1.1 · May 2026
orders-pipeline — Postgres → Camel → Kafka
A senior-Java demonstration of camel-main without Spring or Spring Boot. Atomic-claim concurrency via UPDATE … RETURNING, DLQ-first error handling explicitly written to survive the broker-fully-down case, and a live operator dashboard built in HTMX and Tailwind without a build step. Every architectural decision documented in conventional-commits history. A reviewer can read every line of code in twenty minutes and have a complete model of how it works.
A subscription intelligence service for medspa operators transitioning to longevity clinics. Built and run on a four-agent Anthropic Managed Agent stack — Prospector, Reacher, Scout, and Writer. Twenty-seven newsletter issues published across three publications. Operating proof that one senior engineer with the right AI tooling can build and run a real intelligence business while staying available for senior contract work.
Frontier research from longevity-biotech — clinical readouts, mechanism explainers, deal flow. Where the science actually lives, not where the marketing department wants it to.
Practical AI-agent stacks for B2B services. Two-AI workflows, MCP, real builds with conventional-commits history. Not vendor pitch — actual reproducible practice.
Investigative pattern analysis of organized fraud rings, senior-targeted scam infrastructure, and the structural reasons certain frauds keep recurring.
Available now for contract or full-time roles, advisory engagements, and direct system-build work.
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.
Hybrid-being engagements
AI agent & MCP advisoryAnthropic Managed Agents, MCP servers, multi-agent orchestration. Available to design or build similar systems on contract.
Two-AI workflow consultingHow Cowork + Code actually work together in practice. Process design, attribution, code review, the discipline that prevents drift.
Cross-domain system buildsWhen a project crosses engineering, intelligence, publishing, or pattern-analysis lines, Leonard works natively across all of them.