AI architect and senior full-stack developer building AI systems and tools
I work across Go services, distributed systems, TypeScript apps, and LLM workflows. My current focus is practical AI architecture: turning models, tools, memory, guardrails, and developer workflows into software that can be reviewed, operated, and improved.
| Focus | What I Build |
|---|---|
| AI systems | Claude/Codex workflows, MCP tools, memory systems, eval gates, and tool boundaries |
| Backend architecture | Go services, distributed systems, API design, security, and production operations |
| Developer platforms | Repository-aware automation, CI/CD, documentation-as-context, and AI-assisted delivery loops |
| Web products | TypeScript, React/Next.js, Vue, and production-facing platform work |
- Dense-Mem - durable graph memory for LLM hosts, with typed claims/facts, embeddings, team/profile isolation, and MCP access.
- agentool - Vercel AI SDK tool suite for production-ready file, shell, search, web, memory, and compaction tools.
- GitVibe - maintainer-gated AI development pipeline for GitHub issues, discussions, labels, workflows, branches, and pull requests.
- VCP - security-first protocol for AI-generated code, reviewable guardrails, and multi-AI pipeline orchestration.
- git-doc-mcp - YAML-defined MCP server pattern for exposing repository documentation to AI assistants.
My website is the source of truth for project details, links, and status:
View my projects on markhuang.ai
| Area | Tools |
|---|---|
| AI architecture | Claude, Claude Code, Codex, MCP, Vercel AI SDK, agentic workflows |
| Languages | Go, TypeScript, C#, Python, SQL, JavaScript, Bash |
| Frontend | Next.js, React, Vue.js, HTML, CSS, Sass |
| Backend | Go APIs, .NET, Chi, GORM, REST, GraphQL |
| Infrastructure | Kubernetes, Docker, GitHub Actions, GitLab CI, Redis |
| Data | PostgreSQL, SQL Server, Neo4j, JSON, YAML, Markdown |
I write at markhuang.ai/blog about AI systems, Go services, distributed systems, LLM workflows, AI memory, multi-AI review, prompt/control-plane design, and AI safety guardrails.
Recent themes:
- durable AI memory beyond vector search
- system prompts, user prompts, and control-plane boundaries
- multi-AI review workflows and deterministic gates
- practical software lessons from production systems
- AI architecture, LLM systems, fine-tuning, inference efficiency, and agentic systems
- motorcycles, including casual riding on a 2025 CBR650R
- reading technical books, science fiction, and history
- learning new tools deeply enough to use them well, without pretending every user needs to become a toolmaker
