Archibald大梁

Full-stack engineer · AI-native · 6 yrs

I build backends and frontends, and these days I spend most of my time making LLMs actually finish the job.

Six years end to end: Java/Spring services, PostgreSQL and data pipelines, React/TypeScript interfaces. The past two years have gone AI-native — tracking model releases, agent tooling, RAG, and turning “AI can do the work” into something that ships.

The Roche-limit era of technology技术的洛希极限时代

Selected work

open source · 2
  • Silicon River2025 — ongoing

    Turning the daily flood of model releases into a timeline you can watch, click and query

    Python · FastAPI · PostgreSQL · React · Three.js · TypeScript · DeepSeek · Docker

  • Idea Probe2026 — MVP

    An idea engine on my own machine — Reddit threads through an LLM, keeping only claims and evidence

    Python · DeepSeek · Reddit API

Other projects

  • NexmarkA SaaS marketing system built for WM: Spring Boot 2.6 + PostgreSQL + MyBatis-Plus + Druid server with BCrypt hashing and full DDL scripts, plus a separate portal frontend.
  • Mind CardUmi Max frontend with a Java server — a full-stack exercise across the whole chain.
  • Xmoment OfficeA React + Vite portal frontend. The Vercel demo URL recorded on the repo now 404s, so only the source remains.
  • codegAggregates AI coding sessions from Claude Code, Codex, OpenCode and others into one workspace. Rust, shipping as desktop app, self-hosted server or Docker. I forked it to follow the multi-agent collaboration design. [fork]
  • SnapzyA native macOS screenshot and screen-recording app, an open-source CleanShot X alternative written in Swift. I follow it for how native UI gets built. [fork]

What I do

AI-native apps

LLM application design, cross-validating multiple data sources, agent tooling, RAG and prompt engineering — pushing past “it runs” to “it can ship”.

  • LLM application design
  • DeepSeek
  • Web-augmented retrieval
  • Multi-source cross-validation
  • RAG
  • Agent tooling
Backend & data

Service APIs, data pipelines and caching layers; the kind of work that means chewing through data, static analysis and lineage graphs.

  • Java
  • Spring Boot
  • MyBatis-Plus
  • Python
  • FastAPI
  • PostgreSQL
  • Redis
  • Druid
Frontend & visualization

Building interfaces with real complexity, including 3D data visualization and filter-heavy dashboards.

  • TypeScript
  • React
  • Three.js
  • Vite
  • Umi Max
  • Tailwind
Engineering & delivery

Containers, automated deploys, test coverage and observability; I like to know the boundaries and the degradation path before writing the happy path.

  • Docker
  • docker-compose
  • nginx
  • GitHub Actions
  • Vercel
  • pytest
  • Moving this blog from Hexo to Astro

    The old blog only left behind compiled HTML — the source was gone. The migration had to do three things at once: let the new site take the root path, break zero old links, and stay reversible.

Contact

Good roles, collaborations, or just a conversation about AI engineering — email is the fastest way.