Hi,I'm YIXIN Yuanyixin Chai嗨,我是忆鑫 柴缘忆鑫
I spent the past several years turning simulation and machine-learning data into practical software—from Qt desktop systems and model-training interfaces to reliable data applications. Today I am channeling that engineering foundation into AI agents and open-source developer tools, approaching AI as a disciplined apprentice: learning in public, testing assumptions, and building small systems that earn trust through evidence.
过去几年，我将仿真与机器学习数据转化为可落地的软件，覆盖 Qt 桌面系统、模型训练界面与可靠的数据应用。如今，我正将这套工程基础投入 AI Agent 和开源开发者工具，并以严谨的学徒心态学习 AI：公开实践、验证假设，用证据逐步构建值得信赖的小型系统。
Education教育经历
Work Experience工作经历
What I Build我的工作
Simulation & Desktop Systems仿真与桌面系统
C++ and Qt applications for real-time communication, scenario visualization, model configuration, service monitoring, and operator-focused technical workflows.
使用 C++ 与 Qt 构建实时通信、场景可视化、模型配置、服务监控及面向操作人员的技术工作流。
Real-Time Communication Systems实时通信系统
Point-to-point relay modules that dynamically assign the nearest available source, optimize signal paths, and maintain low-latency connectivity across changing scenarios.
构建点对点通信中继模块，动态分配最近可用信源、优化信号路径，并在变化场景中保持低延迟连接。
Trajectory Model Training Tools轨迹模型训练工具
Qt interfaces for trajectory dataset management and RNN, CNN, LSTM, and GNN parameter configuration, integrated with a Python backend for model training and prediction output.
使用 Qt 构建轨迹数据管理及 RNN、CNN、LSTM、GNN 参数配置界面，并集成 Python 后端执行模型训练与生成预测结果。
Service Monitoring Platform服务监控任务管理平台
A centralized Qt platform for managing network services and tasks, with batch or individual start, stop, and edit controls, parameter-driven configuration, and QtCharts operational visibility.
为网络服务与任务构建集中式 Qt 管理平台，支持批量或单独启动、停止和编辑，并通过参数化配置与 QtCharts 提升运行状态可视性。
Engineering Toolkit技术工具箱
Open Source Project开源项目
I build focused AI systems and developer tools with explicit boundaries, reproducible evaluation, and production-minded engineering.
我专注构建边界清晰、评测可复现，并兼顾生产实践的 AI 系统与开发者工具。
Production Knowledge Research Agent
A production-oriented research agent that combines PostgreSQL and pgvector knowledge retrieval with live web research to produce evidence-grounded, cited answers.
面向生产实践的知识研究智能体，融合 PostgreSQL、pgvector 本地检索与实时网络研究，生成基于证据且带引用的回答。
AI Resume Optimizer
An evidence-grounded CLI that analyzes PDF or DOCX resumes against a job description and produces validated Markdown and editable DOCX outputs without inventing facts.
基于证据的简历优化 CLI，对照职位描述分析 PDF 或 DOCX 简历，在不虚构经历的前提下生成经验证的 Markdown 与可编辑 DOCX。
Context Window Compressor
A provider-neutral Python library that manages agent message history, preserves tool-call groups atomically, applies replaceable compression strategies, and reports detailed metrics.
与模型服务商无关的 Python 库，用于管理智能体消息历史、原子化保留工具调用组、替换压缩策略并输出详细指标。
AI GitHub Reviewer
A read-only single-agent CLI that retrieves a public pull request through a strict GitHub tool, validates the generated review, and never writes back to GitHub.
只读的单智能体代码审查 CLI，通过受限 GitHub 工具读取公开 PR，验证生成的审查结果，并且不会向 GitHub 写入内容。
Prompt Engineering Workbench
An auditable workbench for six deterministic or seeded prompt ablation variants, bounded agent execution, reproducible result persistence, analysis, and reporting.
用于六种确定性或带随机种子的提示词消融实验，涵盖受控智能体执行、可复现实验结果持久化、分析与报告。
LLM Context Explorer
An offline-capable CLI for inspecting message-level token estimates, cumulative context usage, input capacity, limit status, and machine-readable JSON reports.
支持离线运行的上下文分析 CLI，可检查消息级 Token 估算、累计用量、输入容量、窗口状态并输出机器可读 JSON 报告。