CYNTHIA WONG

I'm Cynthia Wong, a Data Science and Big Data Technology student in a Sino-American joint degree program.

Learn more
01 — Education

My Background

NSU NSU

Northeastern State University (USA)

GCTB GCTB

Guangzhou College of Technology and Business
(CN · GCTB-NSU)

Data Science And Big Data Technology | 数据科学与大数据技术 (中外合作办学)
Sino-American Joint Degree Program | 2022.09–2026.07（中国教育部、中留服认证）
Bachelor of Engineer (CN) | Bachelor of Technology (U.S)

3.60 GPA / 86

Main Courses（Includes certificate of instruction in English）:
Data mining(96)、Business statistics(96)、Object Based Visual Program(95)、Machine Learning Application(95)、BigData RealTime Analysis、Database management systems、Data telecommunication，Intro to information security，Html5&CSS、C/C++ programs、concepts of industry，Linear Algebra、Probability and Statistics...etc.

👾
Tech Stack
AI platforms, programming languages & analytics tools
Claude Desktop + CLI Codex Google AI Studio Perplexity Dify Platform Python HTML & CSS BI MATLAB C/C++ Wireshark WEKA R
cynthiawongwingsaan@gmail.com
02 — The Toolkit

Business & Creative Stack

AI-powered, design-aware, data-driven — the tools and craft I bring to every project.

🎨 Design
Creative Tools
Full-stack design workflow from wireframe to production
Figma Canva OpenDesign
📊 BI & Data
Business Analytics
Business intelligence & data storytelling with modern BI tools
Looker Studio Semrush Google Trends
🎬 Video
Content Creation
Video editing & multimedia production for social & academic
Capcut 剪映
🔍 SEO
Growth Marketing
Programmatic SEO pipeline, keyword strategy & technical audit
WordPress Kadence Excel
Languages
Cantonese
Native
Mandarin
二甲 (Professional)
English
CET4 · Daily Communication
03 — Internship

Work Experience

Specialist — scaled content platforms with AI-driven programmatic SEO & multilingual growth strategies.

GeeLark — Google Programmatic SEO Intern
📍 广州磁湖信息技术有限公司 📅 2025.06.30 – 2026.05.30
1200+ Pages Live
3.44M+
Search Impressions
44.5K+
Organic Clicks
19
Avg. Ranking Position
  • Built RAG-based programmatic content generation & publishing Pipeline↗ (Python / 0-1 portfolio piece) — 1 page indexed by OpenAI crawler. Workflow synced to Dify platform.
  • Iterative Optimization & Scalable Execution of Two GEO Pipelines: Took over and managed two Python-based programmatic SEO pipelines for the team, specifically focusing on automated Glossary Generation and Multilingual Blog Translation.
    YOLOv3 SQLite NLTK langdetect Janome fuzzywuzzy mistune ThreadPoolExecutor LLM Quality Control Loop
  • Keyword Research & Intent Analysis: Proficient in leveraging industry-standard platforms like Semrush and Google Trends for keyword discovery and competitive analysis. Integrated AI tools to parse search intent and systematically expand long-tail keywords.
  • Technical SEO Auditing: Conducted comprehensive site-wide audits to identify and backtrack 404/403 internal links for URL remediation. Audited external link status codes (404/403/502) and successfully troubleshot various indexing anomalies.
Performance Report (Click to enlarge)
Performance Report
04 — My Projects

Featured Work

Four projects spanning AI, data engineering, and innovation.

2025.06 – 2026.05 · GeeLark · SaaS

Programmatic SEO Pipeline (0-to-1 Portfolio)

End-to-end AI content generation pipeline — from web scraping to WordPress publishing.

  • Used Playwright + playwright_stealth for anti-detection browser automation and batch asynchronous web scraping; paired with BeautifulSoup for HTML cleaning and core text extraction.
  • Implemented language detection, automatic translation, and ASCII ratio checks to trigger en_model translation; integrated Google Search API for keyword SERP data collection.
  • 9-Step Workflow & LLM Orchestration: Raw data scraping (80%) → Auxiliary page search (20%) → Intelligent FAQ filtering → Product info fusion & optimization (Product Info Base) → AI data cleaning & structure generation (Prompt Engineering via LLMs like Gemini-flash/GPT-4o) → B2B business copy polishing → English synonym replacement → Output format standardization → Excel variable table export.
  • Built 8 image templates using Pillow; achieved input-to-mockup programmatic output via Playwright auto-screenshot, smart cropping, and layer blending.
  • Constructed smart variable replacement flow based on openpyxl + regex (re) to replace placeholders with batch-generated content; output WordPress Kadence Blocks HTML for bulk publishing.
🕷️
1 Page indexed by OpenAI crawler
Python asyncio LLM Orchestration Retrieval-Augmented Generation multi-model routing model registry prompt engineering JSON anti-hallucination guardrails text spinning content deduplication Playwright (async+sync) playwright-stealth BeautifulSoup4 Chromium headless web scraping pipeline GoogleSearch API Pillow (PIL) LANCZOS resampling RGBA compositing batch image processing pipeline openpyxl Markdown HTML
Dify Dashboard — Live Pipeline Preview (Click to enlarge)
Dify Dashboard Preview
2026.02 – 2026.04 · Academic · Graduation Project & Thesis

Facial Expression & Emotion Analysis Model: Design & Implementation (PyTorch + CNN)

  • Designed Lightweight CNN: Built a 4-layer cascaded convolutional network (64→512 channels); applied AdaptiveAvgPool2d + dual Dropout (0.4/0.3), reducing classifier dimensions from 4.6K to 512 to enhance cross-subject generalization capability.
  • Resolved Classification Collapse: Conducted gradient flow analysis and switched to CrossEntropyLoss combined with ReduceLROnPlateau dynamic scheduling; achieved a Weighted F1 score of 0.6935.
  • Built Composite Data Augmentation Pipeline: Utilized RandomAffine + RandomHorizontalFlip + RandomErasing to enable the model to learn distributed texture representations across muscle regions.
  • Automated Experimental Reporting Loop: Leveraged openpyxl + matplotlib to dynamically export multi-sheet training data, curves, and architecture diagrams, achieving end-to-end automation from training to reporting (train → report).
F1 Score 0.6935
Python ReLU PyTorch Conv2d Convolutional Neural Network(CNN) BatchNorm2d MaxPool2d Softmax AdaptiveAvgPool2d Dropout CrossEntropyLoss Adam Optimizer ReduceLROnPlateau Learning Rate Scheduler Data Augmentation RandomAffine RandomHorizontalFlip RandomErasing Grayscale Normalization RMSE Stratified Sampling torchvision PIL/Pillow NumPy pandas scikit-learn Precision Confusion Matrix matplotlib openpyxl
model.py train.py
Code loading... Place your .py files in assets/code/ folder.
2025.05 – 2025.06 · Personal Project

Spotify Music Analytics — Personal Data Portfolio (0-1)

  • Accessed Developer Dashboard & Invoked Spotify API: Configured OAuth redirect URI for user authentication; scraped custom playlists and all track information.
  • Parsed Personal Spotify Extended Streaming History: Conducted data engineering via Pandas; utilized Matplotlib and Seaborn for data visualization and analysis.
  • User Skip Behavior Prediction: Built a K-Nearest Neighbors (KNN) model to predict whether a user skips a track; utilized Confusion Matrix for performance evaluation, analyzed feature correlations using Pearson Correlation Coefficient, and plotted 2D scatter plots; achieved 80% accuracy with strong classification performance.
  • Cross-Platform Playback A/B Testing & Playback Termination Prediction: Performed independent samples t-test and chi-square test on platform groupings (p-value = 0.0568). Implemented Random Forest and KNN multi-class classification models; Random Forest achieved 85% accuracy. Utilized Confusion Matrix and classification reports to identify mainstream categories, achieving F1-scores of 0.91 and 0.95.
Accuracy 80% (KNN)
Accuracy 85% (RF)
F1 0.91–0.95
Python Jupyter Notebook pandas numpy scikit-learn KNeighborsClassifier RandomForestClassifier StandardScaler LabelEncoder scipy.stats ttest_ind chi2_contingency matplotlib seaborn heatmap boxplot barplot scatter confusion_matrix spotipy SpotifyOAuth Spotify Web API
Skip Prediction (KNN) A/B Test Platform
Code loading... Place your .ipynb files in assets/code/ folder.
2023.03 – 2023.07 · Competition

2023 China College Students' Innovation Competition (International Track + Domestic Higher Education Track)
中国大学生互联网+创新创业大赛 (国际赛道+国内高教主赛道)
智能农业植保无人机立项

  • Core Team Member: Led project topic selection and constructed the overall project framework.
  • In-Depth UAV Industry Research: Reviewed 20+ industry reports to gain deep insights into market demands and technical gaps within the "low-altitude economy development" and agricultural plant protection sectors.
  • Business Plan (BP) Writing: Authored the executive summary/project overview section of the business plan, comprehensively and precisely articulating key elements including project background and target markets.
Market Research Business Planning UAV Industry Team Leadership
📊 Presentation Deck (Show highlights only)
Slide 1
1 / 13
05 — Campus Leadership

Campus Experience

Student leadership, event management & bilingual hosting.

4 Campus Projects — 0 to 1
Core Member · Party and Youth League Management Office, College Youth League Committee
Independently launched 4 campus-wide events from scratch: departmental basketball tournament, red-culture field research, student congress documentation, and league study planning. Served nearly 400+ students — projects still running today.
Managed Sports & Arts Department and Planning & Liaison Department. Handled end-to-end event operations: planning, budgeting, promotion, execution, and post-event reporting.
Event Management Budget Planning Team Leadership Cross-departmental
Bilingual Host & English Voice Acting
Served as the college's bilingual host for official events and ceremonies. Participated in the 'Youth Talk' project as an English voice actor — a program that won 2nd Prize in Guangdong Provincial University Ideological & Political Work.
Public Speaking English-Chinese 🏆 Guangdong Provincial Second Prize