仓库/Skill 库

混合 RAG 系统，具备 reranking、基于引用的 grounded citations，以及可选 provider-backed 执行的可确定离线评估分级。A hybrid RAG system with reranking, grounded citations, and deterministic offline evaluation tiers with optional provider-backed execution.

有界研究 Agent：plan → retrieve → critique，具备步骤预算、执行轨迹、引用追溯与自由路径 UI，无需 API key。Bounded research agent: plan → retrieve → critique, step budgets, traces, citations, free-path UI. No API key.

A/B 测试结果分析，涵盖整体指标评估与精细用户分群。项目结合了变更对结账漏斗各阶段影响的研究、统计假设检验以及不同群体的关键指标评估。Analysis of A/B testing results, covering both an overall assessment of metrics and detailed user segmentation. The project combines research into the impact of changes on the stages of the checkout funnel, testing of statistical hypotheses, and evaluation of key metrics across different groups

FraudBench-Hybrid 是一个面向时序银行欺诈检测的可复现研究框架，仓库包含数据集配置、防数据泄露的 Heavy Hybrid 集成实现、实验评估、结果、可视化分析及未来研究扩展资料FraudBench-Hybrid is a reproducible research-based framework developed for temporal bank fraud detection. This repository will include dataset configuration, leakage-controlled Heavy Hybrid ensemble implementation, experimental evaluation, results, visual analysis, and material for future research extensions.

面向有据可循的 Agent 工作流的 TypeScript 参考实现，支持混合检索、人工审核与可回放的审计追踪。TypeScript reference implementation for grounded agent workflows with hybrid retrieval, human review, and replayable audit trails.

涵盖 RAG、AI Agent、评估与商业智能的实战 AI 应用项目集。Practical AI application projects across RAG, AI agents, evaluation, and business intelligence.

仓库包含文章与 Excel 文件，收录系统综述《人工智能方法在实体瘤疗效评价标准中的应用》中的 CLAIM 与 FUTURE-AI 评估。Repository containing articles and Excel files with CLAIM and FUTURE-AI assessments for the systematic review Artificial intelligence approaches to the Response Evaluation Criteria in Solid Tumors: a systematic review.

多阶段气候声明检索与排序：BM25、稠密 ANN、融合、重排序与评估Multi-stage climate claim retrieval and ranking: BM25, dense ANN, fusion, reranking, and evaluation

这个仓库展示了我的技能、项目，以及从数据工程转型为数据科学家角色的持续学习历程。This repository showcases my skills, projects, and continuous learning journey as I transition from Data Engineering to a Data Scientist role.

Shadow / Unmanaged API 的发现、分类、安全验证与可复现研究基准。Shadow / Unmanaged API discovery, classification, security validation, and reproducible research benchmarks

面向系统综述数据抽取、证据定位与偏倚风险评估的冻结式（frozen）防泄漏基准。Frozen, leakage-aware benchmark for systematic-review data extraction, evidence localization, and risk-of-bias support

🚀 使用强化学习优化半精度通用矩阵乘法（HGEMM）CUDA kernel，性能超越 cuBLAS 及其他基准。🚀 Optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels using reinforcement learning, surpassing cuBLAS and other benchmarks with superior performance.

Islam West Africa Collection 的 RDF 词汇表：面向 Omeka S 的 AI 处理溯源与按模型键控的情感标注属性RDF vocabulary for the Islam West Africa Collection: AI processing provenance and model-keyed sentiment annotation properties for Omeka S

对伊斯兰西非文献集（IWAC）语料库情感分析的交互式可视化，对比 ChatGPT、Gemini 与 Mistral，支持多语言与高级筛选。Interactive visualization of sentiment analysis on the Islam West Africa Collection (IWAC) corpus, comparing ChatGPT, Gemini, and Mistral with multilingual support and advanced filtering.

Agentic RAG：回答前对每条引用与原文进行核对，证据不足时弃答。具备自纠正检索循环、确定性引用锚定以及 prompt injection 防御能力。Agentic RAG that verifies every citation against source text before answering, and abstains when the evidence does not hold. Self-correcting retrieval loop, deterministic citation grounding, prompt-injection defence.

RAMR —— 检索增强记忆可靠性：面向 Agentic-RAG / 记忆系统的抗污染合成基准（附方法与发现）。RAMR — Retrieval-Augmented Memory Reliability: a contamination-resistant synthetic benchmark for agentic-RAG / memory systems (findings + method)

系统综述的主动学习摘要筛选。在全部 26 个 SYNERGY 数据集上基准测试：WSS@95 均值 64.1。Active-learning abstract screening for systematic reviews. Benchmarked across all 26 SYNERGY datasets: mean WSS@95 of 64.1.

本地 LLM 基准测试、RAG、语音交互与 AI 助手Local LLM benchmarking, RAG, voice interaction and AI assistant

🎨 利用 AI 实时生成高保真 UI 设计，对比多版本方案，并跨多个模型导出可直接使用的代码。🎨 Generate high-fidelity UI designs in real-time with AI, compare variations, and export ready-to-use code across multiple models.

将问题框架视为计算状态的可复现研究框架，包含理论、算法与基准，用于内生探究。A reproducible research framework for treating problem frames as computational states, with theory, algorithms, and benchmarks for endogenous inquiry.

通过分析 trace 诊断与调试 RAG 流水线，定位并修复影响答案质量与相关性的问题Diagnose and debug Retrieval Augmented Generation pipelines by analyzing traces to identify and fix issues affecting answer quality and relevance.

可复用的 Python 与 notebook 工作流，用于多语言科研数据准备、统计分析及 AI 文本检测器评估。Reusable Python and notebook workflows for multilingual research-data preparation, statistical analysis, and AI-text detector evaluation.

面向 AI coding agents 的 token 预算 context pack，基于编译器解析的 Kotlin 结构（Analysis API/PSI）构建，并附带衡量其是否优于 chunk RAG 的 benchmark。Token-budgeted context packs for AI coding agents, built from compiler-resolved Kotlin structure (Analysis API/PSI) — with the benchmark that measures whether it beats chunk RAG