AY

袁艺 Alan Yuan

@LikeDreamwalker

AI Agent / Full-Stack Engineer
Hebei Normal University of Science & Technology (2017 - 2021)
Bachelor of Computer Science
Beijing or Remote

AI Agent Engineer focused on long-term memory and context systems for personal agents. Background spans memory visualization (Plaud), cross-platform architecture and generative UI for ambient agents (Ambi), and Previously — an independent open-source attempt at in-context self-evolution.

AI Agent · In-Context Self-Evolution & Memory
Previously (Open Source)AI Agent Engineer· Jun 2026 - Present
An open-source project built on my earlier exploration of user memory and agent context at PLAUD.AI. An agent that remembers by when, not by chat thread; episodic memory stored as Markdown slices in the user's own GitHub repo, linked across months through slices, strands, and three timelines. Designed in-context self-evolution — a direction document with falsifiable hypotheses and an evolving previously document — with evidence-anchored sub-agent recall. MIT-licensed with a live demo.
AI + Hardware · Context & Memory
PLAUD.AIFull-Stack Developer· Aug 2025 - Mar 2026
AmbiFull-Stack Developer· Mar 2026 - Jul 2026
Joined PLAUD.AI in August 2025, reporting directly to the China CEO, and helped build Plaud Executive Agent around user context and memory: Mind Palace visualized and made operable the agent's knowledge system; generative node rendering crystallized memory into editable nodes, delivered via an SCR email pipeline; established a five-layer automated testing system (500+ cases).
Then followed that CEO to start Ambi, continuing the Plaud Executive Agent product direction with a focus on the ambient agent. Led the cross-platform full-stack architecture design, and served as the bridge between the engineering team and a US GTM partner, handling cross-timezone, cross-language, and cross-cultural communication; delivered a cross-platform architecture — Next.js + React Router for web, React Native for mobile, Tauri + Rust for desktop, native for watch — and designed a full-stack generative UI system based on AG-UI.
AIGC · Generative Content
HPC-AI TechFrontend Developer· May 2024 - Jul 2025
Frontend team lead. Designed and delivered Video Ocean around the Open-Sora model, covering image/video generation and subscription commerce. Built on RSC hybrid rendering and atomic generation components. Drove the team toward full-stack capability, from 0 to 1.
Internet · User Growth / Monetization
DiDi GlobalFrontend DeveloperUser Growth (UT)· May 2021 - May 2024
Technical Owner for the car-owner monetization direction, covering four business areas and 10+ products. Built deep business experience through commercialization, abstracted business core for logic reuse, sustained rendering and performance optimization, and also contributed to user growth initiatives.
Previously
Jun 2026 - Present
Vercel AI SDK v7Vercel WorkflowEdgeOctokitNext.js 16
An open-source personal AI agent that remembers by when, not by chat thread. Episodic slices + strands link moments across months; memory lives as Markdown files in the user's own GitHub repo — an inspectable, no-lock-in filesystem.
In-context self-evolution: a direction document holds the user portrait and falsifiable hypotheses while the previously document evolves through Past / Now / Horizon; evidence-anchored sub-agent recall with durable runs via Vercel Workflow.
Ambi
Mar 2026 - Jul 2026
TauriRustNext.jsReact NativeAG-UI
Ambi
Cross-platform ambient agent product. Led the cross-platform full-stack architecture: Next.js + React Router web, React Native mobile, Tauri + Rust desktop, native watch — one monorepo sharing components and design tokens.
Designed a full-stack generative UI system on AG-UI + A2UI, with agent streaming events driving dynamic UI and one engine codebase reused across platforms.
Plaud Executive Agent
Aug 2025 - Mar 2026
Three.jsCloudflare WorkersD1AWS LambdaPlate.js
PLAUD.AI
Knowledge space product for an AI hardware ambient agent. Mind Palace visualized and made operable the agent's knowledge system; generative node rendering crystallized memory into editable nodes.
Designed a type-less node system and eight knowledge graph tools for agent Tool Calling. Deployed on Cloudflare Workers and AWS Lambda.
Video Ocean
Jun 2024 - Jul 2025
MDXshadcn/uinext-intl
HPC-AI Tech
Commercial product built around Open-Sora, covering image/video generation, editing and AI chat capabilities.
Designed RSC-based hybrid rendering architecture with granular component decomposition to minimize client runtime overhead.
Introduced atomic generation components decomposing parameters by model/modality dimensions, enabling flexible configuration and rapid reuse.
Frontend team lead, guided the product from 0 to 1 delivery.
DiDi Car Owner Service: Commercial
Mar 2022 - May 2024
MPXVue 2.x
DiDi Global
Four business directions with over ten frontend products, approaching ¥10M in annual profit. Abstracted the shared business core for substantial logic reuse through deep analysis of aligned sub-business processes. Sustained investment in rendering and performance optimization to improve business conversion efficiency.
One key characteristic of commercialization is that all sub-business processes are highly aligned. By abstracting the business core, we can significantly reuse existing business logic, reduce development costs, and thereby improve overall business profitability.