AI Product Manager candidate focused on data quality, evaluation, cloud-native delivery, and agent-ready product loops.

I want to build AI products that solve traditional-industry pain points in real business settings. My bias is toward measurable loops: scenario definition, controllable data generation, benchmark evaluation, cloud-native delivery, and feedback that improves the next iteration.

我的长期方向是：用 AI 解决传统行业的真实痛点，把能力落到具体业务、具体场景和可验证的产品闭环中。

industry pain point
  -> product scenario
  -> controllable synthetic data
  -> benchmark and bias evaluation
  -> cloud-native delivery
  -> agent-friendly documentation
  -> product loop improvement

• Data quality - Synthetic data should be measurable, steerable, and auditable instead of just plentiful. Proof surface: quality distribution control, retention gates, evidence ledgers.
• Education evaluation - AI education products need repeatable benchmarks, not isolated demos. Proof surface: leaderboard, bias analysis, held-out tutoring transfer, process evaluation.
• AI infrastructure - Product ideas fail if runtime, routing, observability, and recovery are vague. Proof surface: AI infra work at Seeles, Skill/MCP routing, request-runtime discipline.
• Cloud native - AI products need production surfaces that operations teams can trust. Proof surface: current Tencent cloud-native business, Tencent Cloud/TCCLI, and TKE tooling.
• Agent documentation - Docs should become executable context for people and agents. Proof surface: loop engineering, SSOT-oriented adapters, CLI-first guides.

• Tencent - Cloud-native business, developer workflows, cloud operation surfaces, and automation.
• Seeles - AI infrastructure related business, with attention to model/application runtime reliability.
• Research track - Synthetic-data quality control, education benchmark design, evidence packaging, and paper-facing analysis infrastructure.

• Education benchmark and evidence infra - Converts AI-education claims into comparable leaderboard, bias, transfer, and process-evaluation surfaces.
• Controllable synthetic-data quality framework - Studies how to make synthetic-data quality distributions more controllable and useful for downstream tasks.
• Tencent Cloud TCCLI Skill - Standardizes Tencent Cloud API operations through tccli, making cloud workflows more agent-operable. Repo
• TKE CLI and workshop materials - Packages cloud-native operating knowledge into repeatable guides, workshops, and CLI-first workflows. Guide / Workshop
• cc-switch-adapter - Routes skill and MCP operations through a centralized SSOT to prevent scattered agent-tooling state. Repo
• Image2PPT - Converts courseware screenshots into editable PPT through image layering, super-resolution, OCR, and VLM reasoning. Repo
• QA from PDFs - Builds validated QA JSONL from paired question and answer PDFs for education datasets and benchmark construction. Repo

• Master's degree - Shaanxi Normal University
• Bachelor's degree - Chengdu University of Information Technology

I am open to AI product roles where product judgment, AI evaluation, data infrastructure, and industry landing all matter.

• Email: kerwin01130224@gmail.com
• GitHub: Kerwin0224

_{Profile README maintained in Kerwin0224/Kerwin0224.}