Hello.
I'm Jemy Yan, an AI Agent Product Engineer based in China. I turn complex agent
capabilities into workflows that professional users can control, trust, review, and improve over time.
At MoTA, I work on an AI investment agent platform for research automation, strategy analysis, risk support,
and investment decision workflows. I focus on turning LLM capabilities into usable product behavior: agent
state, workflow orchestration, evidence tracing, feedback loops, and human-in-the-loop control.
const agentProduct = {
domain: 'investment research',
workflow: ['analyze', 'explain', 'approve', 'learn'],
guarantees: ['controllable', 'auditable', 'feedback-aware']
};
About
I specialize in building production-grade AI agent products for knowledge-intensive work. My background spans
frontend, client-side, full-stack delivery, and AI integration. Today, I focus on the product engineering layer
around agents: how they plan, request human approval, expose uncertainty, cite evidence, recover from mistakes,
and learn from user feedback.
For vertical agents, I believe durable value comes not only from execution, but from memory, workflow fit, and
the ability to model how users and teams make decisions over time.
Product Engineering Focus
Agent Workflows
Designing task states, multi-step flows, tool usage, review checkpoints, and
recoverable paths for professional agent systems.
Memory and Feedback
Designing memory beyond RAG: time-aware, evidence-backed systems that can update,
expire, handle contradictions, and improve from real outcomes.
Trustworthy Interfaces
Shaping probabilistic AI behavior into interfaces with explanation, traceability,
human control, risk visibility, and audit-ready interaction histories.
Skills and Expertise
AI Agent Product Engineering
Agent Workflow Orchestration
Human-in-the-loop Systems
Agent Memory Design
RAG Systems & Vector Databases
LLM Integration (OpenAI / Claude / Gemini)
Prompt Engineering & Evaluation
Explainable AI Interfaces
Financial Research Workflows
Full-Stack Development
JavaScript / TypeScript / Python
Vue.js / Nuxt.js / Node.js / Flutter
AI-Assisted Development (Claude Code / Codex)
Work and Projects
I build AI-powered systems where model output must
become dependable product behavior: observable, explainable, correctable, and useful inside real user
workflows.
Current Work
MoTA is an AI investment agent platform focused on research automation, strategy
analysis, risk support, and investment decision workflows. My work sits at the intersection of agent
capability and product experience: workflow design, agent states, human review, evidence tracing, and feedback
mechanisms for professional users.
Open to Opportunities Worldwide
Interested in roles in vertical agents, memory systems, human-in-the-loop
workflows, AI product engineering, and professional decision-support tools. Open to remote or on-site
positions worldwide. Preferred locations: Hong Kong, Singapore, Dubai, Toronto.