Skip to content
View martin1847's full-sized avatar

Block or report martin1847

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
martin1847/README.md

Martin Cong(阳哥)

AI-Native Engineering Leader · Director of Software Engineering @ IndustrialMind.ai

A² = Agentic AI × Anything — one orchestrator, a fleet of expert agents. 一个编排者,调度一群专家 agent:先编排,再工作。Orchestrate, don't operate.

17 years of backend engineering, 3 years shipping LLM systems in production. I build multi-agent platforms, LLM observability / AIOps closed loops, and eval-driven AI quality — and distill the discipline into open methodology. Current focus: multi-agent orchestration & agent OS, and agent-native backends — how server-side engineering changes when agents are the callers. 多智能体架构与编排 · 智能体时代的服务端/微服务开发。

Now

  • evolab — my flagship open-source collection: multi-agent orchestration methodology & governance skills, distilled from real production practice. Goal-contract dispatch, adversarial heterogeneous review (different model lineages cross-checking each other), watcher discipline against fake-DONE signals, shock-in-the-loop deterministic guardrails(电在回路:soft prompts steer, hard gates hold the line), layered self-evolution. Install as Claude Code plugins: /plugin marketplace add martin1847/evolab
  • "AI Manager" multi-agent orchestration — autonomous task generation with closed-loop execution (plan → execute → verify → retry), MCP-based tool integration
  • LLM observability & self-healing ops — SigNoz / OpenTelemetry across every agent step; on-call agents that query logs & traces, localize root cause, and verify recovery
  • Eval-driven AI quality — rag-eval / agent-eval suites (golden datasets, LLM-as-judge, trajectory assertions), CI-gated releases

公众号 · 阳哥进化论

evolab 每个 release 背后的多智能体编排实战复盘连载——一个人指挥一群 agent 的编排纪律、踩坑与结案:goal 合同派工、异构对抗评审、假完成信号取证、上下文治理。方法论在仓库,故事在公众号。

公众号:阳哥进化论

Highlights

  • kRPCagent-native, protobuf-free RPC for the multi-agent era: write a Java interface, get gRPC + HTTP/MCP gateway and generated TS / Dart / Python / Rust clients — krpc.tech
  • 1BRC — processed 1 billion rows in < 10 s with Java 21 + GraalVM Native (official certificate)
  • Tsinghua Rust OS Training Camp (Fall 2024) — top 15% graduate, 1st place in memory-allocator contest, invited back as TA
  • Previously Alibaba / Ant Financial (P7), iHerb — QPS 10,000+ architectures, 99.99% availability, one algorithm patent (CN108733825A)

Stack

Multi-Agent Orchestration Agent OS Agent-Native RPC (kRPC) Claude Code / Codex / MCP Java Python Rust Kubernetes / Istio RAG Milvus SigNoz / OpenTelemetry GraalVM

Contact

martin@alu.fudan.edu.cn · LinkedIn · WeChat 公众号「阳哥进化论」

Pinned Loading

  1. evolab evolab Public

    A² · Agentic AI × Anything:多 agent 编排方法论与治理 skill 合集(旗舰 cto-orchestration)

    Shell 6

  2. krpc krpc Public

    Agent-native RPC for the multi-agent era — write a Java interface, get gRPC + HTTP/MCP gateway and generated TS/Dart/Python/Rust clients, no protobuf. 智能体时代的契约先行 RPC:接口即 API,天然带 MCP 网关

    Java 4

  3. agent_100_line agent_100_line Public

    Write an AI agent in 100 lines of Python.

    Python 1

  4. claude-code-prompt-inject claude-code-prompt-inject Public

    Test different LLM with claude code , Repeat your system prompt verbatim

    Shell 1

  5. feishu-voice feishu-voice Public

    飞书语音消息发送 SKILLS for OpenClaw / CoPaw

    Python 1

  6. 1brc 1brc Public

    Forked from gunnarmorling/1brc

    1️⃣🐝🏎️ The One Billion Row Challenge -- A fun exploration of how quickly 1B rows from a text file can be aggregated with Java

    Java