Turn unclear requirements into an actionable plan
Map business processes, set priorities, and define MVP scope and acceptance criteria.
AI-HR · SOP mapping across 8 recruitment stagesAI NATIVE · THINK / BUILD / DELIVERPERSONAL PORTFOLIO / 01
Bringing AI into work,and getting things done.
I'm Zhaoyu / Joel. I connect business problems, product thinking, and AI practice. From frontline product support to building agents and workflows, I care about whether technology can become part of real work and solve concrete problems.
THINK / BUILD / DELIVER
Think → Build → Deliver. Keep moving forward.
BUSINESS × PRODUCT × AIExplore below
Build AI into usable workflows
Connect models, knowledge bases, and tools to build agents, Skills, and application prototypes.
ValueGuard · A three-stage agent decision workflowTurn frontline experience into reliable delivery
Resolve product issues, support training, and document processes and methods in manuals and SOPs.
ClassIn · 26 on-site support engagements and 2 product training sessions01 / SELECTED WORK
All projects Capabilities,backed by evidence.
Each case explains the problem I faced, the decisions I made, and what has been validated so far.
AI applications / 01
Customer signals → Perception / Attribution / Strategy → Human confirmationJOEL / SELECTED WORK
ValueGuard
Evidence-based after-sales decisions, with human oversight
Turn customer signals from multiple sources into priority assessments, explanations of causes, and recommended actions to help triage after-sales exceptions involving high-value customers.
Business & product / 02
Business SOPs → Initial screening workflow → Human reviewJOEL / SELECTED WORK
AI-HR
Understand recruitment first, then choose where AI fits
Organize fragmented recruitment activities into standard tasks, clarifying the handoffs between AI processing, human review, and exception handling.
AI applications / 03
Group questions → Knowledge lookup → Human escalation for complex issuesJOEL / SELECTED WORK
AI Community Support Assistant
From repetitive answers to a support process others can take over
Starting with frequent business inquiries in QQ groups, design a workflow for knowledge lookup, context completion, and escalation of complex issues to a human.
RELIABLE BY PRACTICEReliability comes from
Reliability comes from
delivering, time after time.
In frontline support at ClassIn, I resolved real product issues, helped customers use the product, and turned that experience into materials the team could reuse.
Explore my experience26sessions
On-site product support at universities
2sessions
Product training sessions led independently
60+
Training attendances
FROM WORK TO SYSTEM
On-site support → Manuals and tutorials
Training experience → Standardized SOPs
02 / WRITING & NOTES
All writing After doing the work,think a little deeper.
Deep reflections · Mind map / Feishu original
A Big-Picture Framework for AI(opens in a new tab)
My big-picture, framework-level thinking about AI products, presented as a mind map.
Deep reflections
After Building Many AI Productivity Tools, I Realized I Might Be Solving the Wrong Problems
From local efficiency to business outcomes: define the goal, identify constraints, and use an MVP to test the most critical uncertainty.
Deep reflections
AI Made Me More Capable. Why Hasn't My Reality Changed?
Distinguish creating value from capturing it: once AI is no longer the bottleneck, look again at the variables limiting opportunities, resources, and outcomes.
LET’S CONNECTLet's get to know each other
Contact me Let's get to know each other
through a concrete problem.
Career opportunities, project collaboration, or a conversation about bringing AI into real work.