Dawit Abdisa (Dave) 大卫

Master's Thesis Project 硕士论文项目

Developed an AI-driven decision-support system for PV and battery energy storage feasibility analysis, combining multi-agent workflow orchestration, geospatial and solar resource assessment, scenario-based system sizing, financial modeling, and automated reporting.

开发了一个面向光伏与电池储能可行性分析的 AI 决策支持系统，整合多智能体工作流编排、地理与太阳能资源评估、情景化系统配置、财务建模与自动化报告生成。

• Problem: 问题： Manual feasibility studies were slow and inconsistent. 传统可行性分析流程耗时且一致性不足。
• Approach: 方法： Built a multi-agent workflow with geospatial analysis + financial simulation. 构建多智能体工作流，融合地理分析与财务仿真。
• Outcome: 结果： Generated faster scenario reports for PV-BESS decision making. 显著加快了光储方案的情景报告生成与决策效率。

RAG Assistant RAG 助手

Conversational RAG assistant: upload PDFs, TXT, or Markdown and ask questions grounded in your documents, with source citations and chat memory. Built with DeepSeek, LangChain LCEL, ChromaDB, and Streamlit. Includes RAGAS evaluation harness and LangSmith tracing.

对话式 RAG 助手，支持上传 PDF、TXT 或 Markdown 文档并基于内容进行问答，提供来源引用与会话记忆。基于 DeepSeek、LangChain LCEL、ChromaDB 与 Streamlit 构建，并集成 RAGAS 评估与 LangSmith 链路追踪。

• Problem: 问题： Teams needed reliable Q&A from private docs. 团队需要基于私有文档的可靠问答能力。
• Approach: 方法： Implemented retrieval pipeline with citation grounding and eval traces. 实现带来源引用的检索链路，并接入评估追踪。
• Outcome: 结果： Improved answer trust and observability in production-style workflows. 提升了答案可信度与流程可观测性。

GitHub →

SQL Analyst SQL 智能分析助手

Ask questions about your data in plain English; converts natural language to SQL, auto-selects the best chart type, and returns AI-generated bullet-point insights. Blocks mutating queries for safety. Built with DeepSeek, LangChain, Plotly, and Streamlit.

支持使用自然语言查询数据，自动将问题转换为 SQL、选择合适图表，并生成 AI 分析要点。系统会拦截修改型查询以增强安全性。基于 DeepSeek、LangChain、Plotly 与 Streamlit 构建。

• Problem: 问题： Business users needed analytics without writing SQL. 业务用户希望无 SQL 门槛完成分析。
• Approach: 方法： Natural-language to SQL with chart recommendation and safe query guardrails. 自然语言转 SQL，自动图表推荐，并加入安全查询防护。
• Outcome: 结果： Delivered faster self-serve insights for non-technical stakeholders. 让非技术角色更快获得可执行的数据洞察。

GitHub →

Todo & Email AI Agent 待办与邮件 AI 智能体

Agentic assistant that reads, drafts, and sends emails via Gmail API while managing a to-do list across multi-step reasoning loops. Built with LangGraph, Gmail API, DeepSeek, and Streamlit.

基于智能体的助手，可通过 Gmail API 阅读、起草并发送邮件，同时在多步推理流程中管理待办事项。项目使用 LangGraph、Gmail API、DeepSeek 与 Streamlit 构建。

GitHub →

Prompt Injection Scanner 提示词注入扫描器

CLI tool that tests LLM system prompts against 22 categorized prompt injection payloads using keyword and LLM-judge evaluation modes. Detects jailbreaks, role-play bypasses, and data-extraction attacks.

命令行工具，使用关键词规则与 LLM 评审两种模式，对大模型系统提示词进行 22 类提示词注入载荷测试，可识别越狱、角色扮演绕过与数据提取攻击。

GitHub →

Vulnerable Web App 漏洞教学 Web 应用

Intentionally vulnerable Flask + SQLite web application with 6 documented vulnerabilities (SQLi, Stored XSS, Reflected XSS, IDOR) and a professional security assessment report with PoC exploits and secure fixes.

基于 Flask 与 SQLite 构建的故意脆弱 Web 应用，包含 6 类已文档化漏洞，如 SQL 注入、存储型与反射型 XSS、IDOR，并配套 PoC 利用与安全修复报告。

GitHub →

Network Traffic Analyzer 网络流量分析器

Python CLI tool using Scapy to detect port scans (sliding-window algorithm) and cleartext credential leaks (FTP, Telnet, HTTP Basic Auth, POST bodies) from live interfaces or PCAP files, with colour-coded alerts and JSON reporting.

使用 Scapy 开发的 Python 命令行网络流量分析工具，可从实时网卡或 PCAP 文件中检测端口扫描与明文凭证泄露，并提供彩色告警与 JSON 报告。

GitHub →

EthioMarket EthioMarket 电商平台

E-commerce storefront for authentic Ethiopian products, built in both static HTML/CSS/JS and React + Vite versions with shopping, cart, checkout, and blog flows.

面向埃塞俄比亚特色产品的电商平台，提供静态 HTML/CSS/JS 与 React + Vite 两个版本，包含商品浏览、购物车、结账与博客流程。

Live → 在线访问 → GitHub →

helloworldEdu helloworldEdu 学习平台

Learning management platform built with Next.js, TypeScript, Prisma, and PostgreSQL, with course delivery, enrollment, quizzes, certificates, and Vercel deployment support.

基于 Next.js、TypeScript、Prisma 与 PostgreSQL 构建的在线学习管理平台，支持课程发布、报名、测验、证书与 Vercel 部署。

Live → 在线访问 → GitHub →

Ethioweb.net / Ethiopianweb.net Ethioweb.net / Ethiopianweb.net 社区平台

Community-focused web platform for Ethiopians in China, built with TypeScript and Vite-based frontend tooling to share resources, connections, and useful local information.

面向在华埃塞俄比亚人的社区型信息平台，基于 TypeScript 与 Vite 前端工具链开发，用于分享资源、连接人与提供本地实用信息。

ethioweb.net ethiopianweb.net GitHub →

Midroc Gold Hybrid Microgrid Midroc Gold 混合微电网

Led system sizing, hybrid optimization, and feasibility analysis for a 7MW PV + 2MW BESS + diesel off-grid microgrid designed to improve reliability and reduce diesel use for mining operations.

负责 7MW 光伏 + 2MW 储能 + 柴油机离网混合微电网的系统选型、混合优化与可行性分析，目标是提升矿区供电可靠性并降低柴油消耗。

Roha Health Center PV-ESS-EV Microgrid Roha 医疗中心光储充微电网

Designed an 818kWp rooftop PV + 261kWh BESS + EV charging microgrid for a healthcare facility, including backup strategy and grid-tie control logic for critical medical loads.

为医疗机构设计 818kWp 屋顶光伏 + 261kWh 储能 + 电动车充电微电网，涵盖关键医疗负载的备电策略与并网控制逻辑。

German Square Fast-Charging Station German Square 商业快充站

Led layout design, power distribution sizing, and grid integration analysis for a 960kW commercial charging station with 8 dual-gun 120kW DC chargers serving 100+ EVs daily.

负责 960kW 商业快充站的布局设计、配电容量计算与并网分析，配置 8 台双枪 120kW 直流充电桩，日服务能力超过 100 辆电动车。

Midroc Hospital PV & BESS Backup Midroc 医院光伏与储能备电

Designed rooftop PV layout, 10MWh BESS sizing, and backup power control logic to cut grid costs and strengthen resilience for critical hospital operations.

完成医院屋顶光伏布局、10MWh 储能容量配置与备电控制逻辑设计，以降低用电成本并增强关键医疗场景的供电韧性。

Customized PV-BESS Backup Solutions 定制化光储备电方案

Delivered tailored PV-BESS backup systems for bank branches and residential clients through load profiling, system sizing, proposal design, and reliability-focused planning.

通过负载分析、系统配置、方案设计与可靠性规划，为银行网点与住宅客户提供定制化光储备电系统。

SVM-Based Driver Drowsiness Detection 基于 SVM 的驾驶疲劳检测

Built an ML pipeline for classifying drowsy vs. alert states from EEG brain waves. Compared SVM (98%), LSTM (93%), KNN (87%).

构建基于脑电信号的驾驶疲劳识别机器学习流程，对比 SVM、LSTM 与 KNN 模型，其中 SVM 达到 98% 准确率。

Deep Learning: NanoGPT 深度学习：NanoGPT

NanoGPT training & analysis, Transformer architecture study (GPT, BERT, ELMo).

进行 NanoGPT 训练与分析，并系统学习 Transformer 架构及 GPT、BERT、ELMo 等模型原理。

Web Security Testing Web 安全测试

CSRF, XSS, SQL Injection, SSL/TLS penetration testing labs in Docker-based SEED environments.

在基于 Docker 的 SEED 实验环境中完成 CSRF、XSS、SQL 注入与 SSL/TLS 等 Web 安全测试实验。

Study Abroad Management Website 留学管理网站

Full-stack web application for managing study abroad processes.

用于管理留学申请与流程跟踪的全栈 Web 应用。

Maze Robot 迷宫机器人

Autonomous maze-solving robot with sensor integration. Won 3rd Prize at World Vocational College Skills Competition.

集成传感器的自主迷宫求解机器人项目，获得世界职业院校技能大赛三等奖。

Blockchain Implementation 区块链实现

Distributed systems and cryptographic security protocols.

围绕分布式系统与密码学安全协议的区块链实现实践。

Responsive Web Design Projects 响应式网页设计项目

Modern responsive websites using contemporary frameworks and design patterns.

使用现代框架与设计模式完成多项响应式网站开发项目。