01
Building
Evidence Graph
Turns search results into reviewable claims, evidence, and source relationships instead of another untraceable summary.
- Next.js
- TypeScript
- Agent
- pgvector
Do traceable citations make AI research easier to review?
Exact match · Paragraph 18
AI product engineering
Ailian
Senior frontend engineer expanding into full-stack and Agent engineering
I turn complex AI workflows into products that remain understandable, verifiable, and maintainable. My current focus is Evidence Graph: bringing research claims back to source evidence.
01 / Work
Selected work
Three projects under real development, with no placeholder demos.
02
In development
AI Photo Studio CN
An AI photo studio for Chinese portraits, avatars, and social content, focused on complete jobs, provider safety gates, and failure recovery.
- Next.js
- Prisma
- Playwright
- Image API
03
MVP ready
MCP Guardian
A pre-call policy gateway for MCP tool calls—allow, deny, redact, or require approval—with replayable audits.
- TypeScript
- MCP
- Policy
- Cursor
- Codex
02 / Notes
Engineering notes
Decision evidence, failure paths, and verification methods.
Draft
Why ordinary AI search summaries are not enough
Moving from one-off answers to reviewable, versioned evidence projects.
Research
CyberVerse: between technical prototype and production foundation
Reviewing a realtime digital-human Agent project through code, tests, deployment, and security boundaries.
Draft
Keeping paid AI provider tests bounded
Mock defaults, confirmation tokens, budget limits, and auditable failure paths.
03 / About
Product frontend / Full-stack systems / Agent workflows
Senior frontend engineer expanding into full-stack and Agent engineering