冯靖婉 AI产品经理 / Agentic Engineer Coco Feng AI Product Manager · Agentic Engineer
设计师出身，扎根游戏行业，自驱探索 AI 工程。致力于将前沿技术栈转化为解决实际生产痛点的落地工具。
Designer by training, rooted in the gaming industry, self-driven in AI engineering. Dedicated to translating tech stacks into practical tools that solve real production challenges.
UI/UX 产品设计师
UI/UX Product Designer
MCP与API嵌入集成 · 本地模型
MCP & API Embeddings · Local Models
专注工具落地
Focusing on practical adoption
熟悉制作管线与痛点
Familiar with production pipelines
游戏制作 × 产品设计 × AI 应用探索。在游戏公司负责项目管理的同时，自学并实践 AI 工程，为团队开发了批处理与参考生成等基础提效工具。
具备务实的 AI 技术评估与工具开发能力，理解游戏制作管线中的自动化需求。希望能在这个交叉领域，踏实推动 AI 工具在实际业务中的有效落地。
Game production × product design × AI application exploration. While managing projects at a game studio, I self-taught and practiced AI engineering, building foundational productivity tools like batch processing and reference generators for the team.
Possess practical capabilities in AI tech evaluation and tool development, understanding automation needs in game production pipelines. Hope to steadily drive the effective adoption of AI tools in real-world business scenarios at this intersection.
- 自驱 AI 工程实践：在游戏 PM 职责外自学 AI 技能，探索 Agentic Workflow 与本地模型部署，产出多项面向实际生产场景的提效工具
- 百万字文本批量处理：使用 Google AI Studio 搭建基础工具流程，结合 Claude / Antigravity 对游戏剧本进行大批量文本校对与风险核查，显著提升内容审核效率
- Agentic Workflow 探索：基于 Claude Code 构建基础多步骤 AI 任务流程，尝试将内容生成与数据处理环节自动化
- AI 基础设施搭建：在 Hugging Face 选型并测试开源模型，使用 Ollama / LM Studio 在本地部署并运行推理，配置 Cloudflare 反向代理保障开发测试的网络访问
- 批量图像参考生成：基于 GitHub 开源方案定制集成，为团队提供美术参考图的批量生成工具，分担重复性工作
- 项目管理：负责排期与资源分配，保障近期任务按时交付；协助管理外包供应商流程
- Self-directed AI practice: Learned AI skills outside Game PM duties, exploring Agentic Workflow and local model deployment to build practical productivity tools for real production scenarios
- Million-character text batch processing: Built a basic tooling pipeline using Google AI Studio, then used Claude / Antigravity to batch proofread and conduct risk checks on game scripts at scale, significantly improving content review efficiency
- Agentic Workflow exploration: Built a basic multi-step AI task pipeline using Claude Code to automate content generation and data processing steps
- AI infrastructure setup: Evaluated and tested open-source models on Hugging Face, used Ollama / LM Studio to deploy and run local inference, and configured Cloudflare reverse proxy for stable dev/test network access
- Batch image reference generation: Customized an open-source GitHub solution to provide the team with a batch art reference generator, reducing repetitive work
- Project management: Handled scheduling and resource allocation, ensuring on-time delivery of recent tasks; assisted in vendor management
- 参与《侏罗纪世界：进化3》（已发布）等国际游戏项目 VO/ADR 录音工作
- 严格执行数字资产管理规范，在国际化制作流程标准下保障交付质量
- VO/ADR recording for Jurassic World Evolution 3 (shipped) and other international game titles
- Strict digital asset management within an international game production pipeline
- 端到端设计执行：需求分析 → Figma 高保真 UI → 设计系统搭建
- 与产品经理和工程师协作，输出设计规范文档和部分 PRD
- End-to-end design ownership from requirements through high-fidelity Figma UI delivery; built and maintained design system
- Collaborated with PMs and engineers; produced design specs and contributed to PRD documentation
- 参与品牌焕新与官网重建，协助从策略到视觉执行的落地
- 统筹 Art Vancouver 2022 策展协助与现场执行，协调多方资源
- Contributed to brand refresh and website rebuild — supporting strategy through visual execution
- Coordinated Art Vancouver 2022: curation support, on-site execution, artist and logistics management
尝试构建 Orchestrator + Subagent 架构，将求职流程拆分为信息检索、JD 对比与简历优化等子任务，通过智能体协同完成基础自动化流程。
Attempted an Orchestrator + Subagent architecture, breaking down the job-search process into information retrieval, JD comparison, and resume optimization sub-tasks, using agent collaboration to achieve basic automation.
基于 GitHub 开源方案定制化开发的美术参考批量生成工具。将手动搜图转化为自动化流程，解决团队特定痛点，验证了技术落地的可行性。
A customized art reference batch generator based on open-source solutions. Converted manual image searching into an automated process, addressing a specific team pain point and validating practical application.
搭建小型本地 AI 测试环境：对比测试部分开源模型，通过 Ollama / LM Studio 运行本地推理，并配合 Cloudflare 反向代理解决接口访问限制问题。
Set up a small local AI test environment: benchmarked selected open-source models, ran local inference via Ollama / LM Studio, and bypassed API access restrictions using a Cloudflare reverse proxy.
探索 MCP 工具服务器的实际应用，将文件系统、Web 搜索与 API 嵌入结合引进 Agent 工作流。通过标准化接口调用，尝试拓展自动化脚本的实际执行边界。
Explored practical applications of MCP tool servers, introducing the joint integration of file systems, web searches, and API embeddings into Agent workflows. Standardized interface calls to expand the practical execution capabilities of automation scripts.
欢迎与我交流 AI 产品相关的机会 :D
Open to discussions on AI Products and practical applications.