Building reliable backend systems
and production-ready AI Agents.
Java · Spring Boot · Redis · RocketMQ
Spring AI · RAG · MCP · Agent
I build production-minded backend systems and the agent infrastructure that runs on top of them. I care about correctness under load, clear boundaries, and systems that stay observable long after launch.
- Backend & systems — JVM, high concurrency, I/O, messaging (RocketMQ), caching (Redis).
- AI agents — Spring AI, RAG, MCP, and agent runtimes — shipped, not just read about.
- Judgment — restraint over novelty: one agent until many is justified; measure before scaling.
From the edge in: a typed Client / SDK reaches the API Gateway (auth, rate limits), which hands off to the Agent Runtime. That core is where the real work happens — a Planner (Jev) that decides and scores, a Guardrail that judges before any side effect, Memory that compacts context, and an Evaluator that gates releases. Tools and MCP servers are the only ways out, and every hop is measured by Observability.
A Jev-driven enterprise agent runtime / gateway: single-agent orchestration with guardrails, memory, evaluation and observability baked in.
9 modules · 69 source files · Spring Boot 3.4
A customized Git plugin for Obsidian that solves the nested sub-repository upload problem — the gap the built-in sync never covered.
Obsidian plugin · MIT
Turns your GitHub stars into a personal technical constellation.
Visualization · MIT