Product Manager | Applied AI · FinTech · Web3
I turn complex, data-heavy workflows into products people can inspect, understand, and control—from regulated financial reporting and on-chain systems to teacher-in-the-loop AI prototypes.
Based in mainland China · Open to international opportunities and relocation
- AI Assignment Grading Assistant — explores rubric-first AI preliminary scoring: teachers review and lock the rubric, inspect criterion-linked suggestions, and confirm the final grade. When enabled, submission text, assignment name, and rubric go through a local bridge to the configured provider. AI scoring is not a hosted service; accuracy and time savings have not been validated.
- Muzhi Livestock Teaching Assistant — a teacher-side prototype that turns an original or authorized teaching image, with an optional learning focus, into an editable question draft. The image and prompt/context go to the configured DeepSeek API; teachers review the draft before saving it to the current browser's local question bank. The separate student practice page is not connected to this local bank. For teaching practice only, not animal diagnosis or treatment.
These are product prototypes, not validated classroom deployments. Their problem statements are product hypotheses; user outcomes remain to be tested. The repositories contain public-safe case materials, not the full private product source. No open demo link is included here.
- RWA Investor Data & Reconciliation — synthetic financial data, reconciliation, lineage, and grounded AI explanations.
- DEX Trading Product System Design — quote lifecycle, execution states, user-risk controls, and read-only AI decision support.
- Bitcoin-native Asset Analytics — protocol-event normalization, asset/address analytics, and reviewable risk labels.
- AI-Assisted Influencer Content Operations — a structured workflow from persona definition to a publishing board.
- AI-Enabled Tourism Service System — public-safe reconstruction of multi-party service workflows and delivery constraints.
Each repository distinguishes product hypothesis from evidence, prototype from production, and AI output from human-approved decisions. Look for the user task, workflow, trade-offs, failure boundaries, evidence level, and next validation step.