I build reliable tools for complicated systems.
Agent infrastructure · Distributed backends · Full-stack delivery
Flagship project · Selected work · Engineering profile · All repositories
I'm a software engineer who turns multi-step workflows into systems that are easier to understand, operate, and extend. My work spans AI agent coordination, Java and Spring services, and TypeScript product development.
Most of my current work sits at the boundary between agent coordination and dependable software delivery: preserve state, make failure explicit, and keep people in control of consequential changes.
These are the two repositories that best represent how I design and validate software today.
| Project | What it demonstrates | Inspect |
|---|---|---|
| Agent Relay | Local-first task continuity, explicit checkpoints, safe handoffs, and recoverable execution across user-owned AI agents. | Run the demo · Architecture · Tests |
| Codex Council | Independent advisor runs, anonymous peer review, deterministic aggregation, and bounded debate for consequential decisions. | Workflow · Architecture · Latest release |
- Spring Cloud Microservices — service discovery, gateway routing, load balancing, and independently deployable services with Java 17, Spring Boot, Spring Cloud, and Maven.
- E-learning Platform — an authenticated, SQL-backed course product spanning Next.js, TypeScript, Express, and MySQL.
| Focus | What I bring |
|---|---|
| Agent infrastructure | Explicit checkpoints, safe handoffs, schema-backed contracts, and local-first workflows |
| Backend systems | Java and Spring services, API gateways, service discovery, event-driven design, and SQL persistence |
| Product engineering | TypeScript applications, Next.js interfaces, Node.js APIs, and end-to-end delivery |
| Developer experience | Small command-line tools, clear documentation, predictable configuration, and auditable behavior |
- Make state explicit. Important work should survive a process, tool, or agent boundary.
- Prefer evidence over ceremony. Tests, schemas, and observable behavior carry more weight than confident claims.
- Keep the operating model simple. Every dependency and automation should earn its maintenance cost.
- Design the fallback path. A recoverable failure is part of the product, not an afterthought.
If you're exploring reliable agent workflows, local-first developer tools, or backend architecture, open a focused issue in the relevant repository. Include the use case, constraints, and a concrete outcome—that is the fastest way to start a useful technical conversation.
Agent Relay issues · Codex Council issues · All repositories