Full-Stack AI Engineer · RAG · Agents · MCP

I help startups ship production AI features.

Hi, I'm Yuzhe.

I help startups ship

RAG assistants

to production.

RAG, AI agents and MCP integrations in your web app — full-stack delivery, backed by 10+ years of building backend systems.

TypeScriptNode.jsNext.jsPostgreSQLAWSRAG PipelinesAI Agent SystemsSystem Design

system-delivery.ts

const workflow = createWorkflow("system-delivery");

workflow.bootstrap(
  runtime: "nextjs",
  datastore: ["postgres", "redis"],
  observability: true,
});

workflow.attachAgent("rag-ops", {
  retrieval: "hybrid-search",
  guardrails: ["policy", "budget", "fallback"],
});

workflow.deliver(
  market: "global",
  sla: "high-availability",
  cadence: "continuous-iteration",
});

Service Scope

Five ways to move from an unclear requirement to a shipped system.

Built for founders and independent operators who need one experienced engineer to move a product, internal system, or AI capability forward. Every offer starts with the problem and ends with a concrete next step.

Product delivery

Web App / SaaS Development

You have a product idea or business requirement that needs to become a launchable, maintainable web product.

  • Requirements and core flows
  • Frontend, backend, data, and access control
  • Testing, deployment, and launch support

Operations

Admin Dashboard / Operations System

Your team relies on spreadsheets, manual handoffs, or disconnected tools and needs one operational workspace.

  • Roles, permissions, and workflows
  • Dashboards, forms, and operations console
  • Integrations, audit logs, and handoff documentation

Applied AI

AI Feature / RAG / Agent Integration

You want AI inside a real workflow, with evidence and safeguards beyond a generic chat demo.

  • Use-case and data-source design
  • RAG, agent, or model integration
  • Evaluation, guardrails, cost controls, and observability

Technical direction

Architecture and Technical Consulting

A system faces scaling, reliability, technology-choice, or modernization decisions that need experienced independent judgment.

  • Current-state and risk review
  • Architecture proposal and tradeoffs
  • Phased implementation roadmap and review

Paid discovery

Paid Product Discovery

The idea is not yet specific enough for responsible estimation or development without aligning scope, budget, and expectations.

  • Target user and problem definition
  • MVP scope and priorities
  • Technical path, risks, and delivery estimate
See the full service scope

Latest writing

Real decisions, implementation, and outcomes from independent building.

From SaaS and AI systems to global websites, I document what gets built, what changes, and what stops.