Lin Zhang

• Participated in the development of a course management system, focusing on online learning and teaching.
• Wrote automation tests using Jest framework and contributed to documentation.
• Implemented backend features and integrated a database for user information storage.
• Technologies used: JavaScript, React, Python, Jest, Git.

• Developed an application to recommend movies based on user preferences.
• Integrated APIs to fetch movie data and designed frontend and backend interactions.
• Utilized JavaScript, HTML/CSS, Git for development.

The Umbrella Alert Weather App is a full-stack web application designed to help users stay prepared for weather changes by sending personalized email notifications based on real-time weather forecasts. The goal of this project was to create a simple, reliable tool that ensures users never leave home unprepared — whether it’s for rain, heat, or other weather conditions that could affect their day.

Key Features

• Personalized Weather Alerts:
  Users receive automated email notifications whenever weather conditions meet specific thresholds, such as:
• Rain or thunderstorms in the next 12 hours.
• Extreme heat alerts for sun protection reminders.
• Cold weather alerts for extra layers.

Real-Time Weather Data:

• Integrates with the OpenWeatherMap API to fetch accurate, up-to-date forecasts, including temperature, humidity, and precipitation data.

User-Friendly Interface:

• Simple and responsive frontend UI where users can:
• Enter their location (city or zip code).
• Set custom notification preferences for different weather conditions.
• View a 5-day weather forecast at a glance.

Automated Scheduling:

• Uses backend scheduling to trigger email alerts at set times of the day, ensuring users are notified at the most relevant moment — for example, early morning before they leave for work or school.

Lightweight Data Storage:
No traditional database required. The app uses JSON files to:

• Store user preferences and email addresses.
• Track notification history for reliability and debugging.

Tech Stack & Tools

Frontend:

HTML/CSS – Core structure and styling of the web interface.

JavaScript – Dynamic updates and interactive features.

Bootstrap 5 – Responsive, mobile-friendly design.

Backend:

Flask (Python) – Lightweight web framework for handling routes and logic.

Requests Library (Python) – API calls to OpenWeatherMap for live data.

SMTP & Python Email Libraries – Sending automated email alerts with customizable subjects and content.

API Integration:

OpenWeatherMap API – Provides real-time and forecasted weather data.

Version Control:

Git & GitHub – Code tracking, version control, and collaboration.

Impact & Purpose

I created Umbrella Alert to solve a personal pain point: forgetting to check the weather and leaving without essential items like an umbrella or sun protection. The app evolved into a practical tool that demonstrates my ability to design and deploy end-to-end solutions, integrating:

Full-stack development skills (from UI to backend scheduling).

API usage and data handling for real-world applications.

Automation techniques to improve user experience.

This project reflects my interest in building tools that blend technology and everyday life, creating meaningful solutions that people can rely on.

The World Happiness Report Prediction Model is a data-driven machine learning project that analyzes global happiness factors to predict the Happiness Score of countries. The model leverages real-world datasets to uncover patterns, generate insights, and make accurate predictions, demonstrating the power of machine learning to understand and forecast complex social phenomena.

This project not only enhanced my skills in data science and machine learning, but also taught me how to design a complete ML pipeline — from data collection and cleaning to feature engineering, model selection, and deployment.

Key Features:

Comprehensive Data Analysis:

• Uses historical World Happiness Report data spanning multiple years.
• Includes features such as GDP per capita, social support, life expectancy, freedom to make life choices, corruption perception, and generosity.

Interactive Data Visualizations:

• Built visualizations to explore trends and correlations, including:
• Heatmaps showing feature importance.
• Scatterplots comparing happiness scores between countries.
• Year-over-year changes in happiness rankings.

Predictive Modeling:

• Trains machine learning models to predict the happiness score of a country given its socio-economic factors.

Models used include:

• Linear Regression – baseline model.
• Random Forest Regressor – for improved accuracy and feature importance analysis.
• Gradient Boosting – to handle complex, non-linear relationships.

Performance Evaluation:

• Uses Mean Absolute Error (MAE) and R² Score to evaluate and compare models.
• Iteratively improves performance through hyperparameter tuning.

Scalable, Reusable Codebase:

• Modular structure for easy updates with future datasets or additional features.

Tech Stack & Tools

Programming Language:

• Python – core language for all data science and ML workflows.

Data Handling & Analysis:

• Pandas – data cleaning and manipulation.
• NumPy – numerical operations and matrix handling.

Machine Learning:

• Scikit-learn – for model building, evaluation, and tuning.
• Random Forest, Gradient Boosting, Linear Regression – main algorithms used.

Data Visualization:

• Matplotlib – line graphs and scatterplots.
• Seaborn – heatmaps and advanced statistical visuals.

Workflow & Version Control:

• Jupyter Notebook – iterative analysis and model development.
• Git & GitHub – version control and documentation.

Impact & Purpose

The project aims to identify the key factors that drive happiness worldwide, providing insights into how governments and organizations can focus resources for the well-being of their populations.

From a technical perspective, this project demonstrates my ability to:

• Build end-to-end ML pipelines.
• Work with real-world, messy datasets, performing cleaning and feature engineering.
• Interpret data and translate findings into actionable insights.
• Combine data visualization and machine learning for storytelling and decision-making.

By integrating data science techniques with a socially relevant topic, this project showcases how technology can be applied to understand global issues and support better decision-making.

Player Development & Profile System — Project Description
Overview

The Player Development & Profile System is a data-driven platform designed to help our university baseball team maximize player performance and growth. By leveraging over six years of TrackMan data from practices and games, the system transforms raw numbers into actionable insights for coaches, players, and staff.

This project centralizes historical and real-time performance data into interactive profiles that track player progress, identify strengths and weaknesses, and generate predictive analytics.
The end goal is to empower decision-making for training, recruiting, and game strategy through precise, evidence-based reporting.

My Role — Engineer Lead

As the Engineer Lead, I am responsible for the technical vision, design, and development of the platform.
I work closely with the project manager to align technical implementation with team needs, ensuring smooth collaboration between technical and non-technical stakeholders.

Key responsibilities include:

Designing the overall system architecture, including database schema, backend services, and frontend integration.

Leading technical decisions on frameworks, tools, and scalability solutions.

Overseeing data pipeline development to clean, process, and store large volumes of historical TrackMan data.

Building core features such as:

Automated data ingestion and normalization.

Interactive dashboards for player profiles.

Analytics modules for trend detection and performance tracking.

Mentoring team members on code quality, testing practices, and development workflows.

Problem & Motivation

Our baseball program had six years of high-quality data, but it was fragmented, difficult to access, and underutilized.
Coaches relied on manual reports and spreadsheets, making it hard to track trends or make real-time decisions.

Challenges we faced:

Large, unstructured datasets from multiple seasons and sources.

Lack of a centralized platform for coaches and players to access performance data.

Missed opportunities for advanced analytics like injury risk prediction or matchup insights.

The project was born out of the need to unlock the full potential of our data, giving the team a modern toolset to analyze performance and drive improvement.

Core Features
1. Centralized Player Profiles

Each player has a digital profile that evolves over time, including:

Seasonal averages and career summaries.

Key hitting stats like batting average, OBP, SLG, and exit velocity.

Pitching stats like velocity by pitch type, spin rate trends, and release point consistency.

Fielding performance, including reaction time, range factor, and errors over time.

2. Analytics & Insights

The system goes beyond static stats, providing:

Trend analysis: Detect unusual changes in performance, such as declining fastball velocity or improving launch angles.

Benchmarking: Compare individual players to team or league averages.

Predictive modeling: Estimate future performance and identify injury risks using machine learning.

3. Automated Reporting & Alerts

Weekly reports generated as PDFs and delivered directly to coaches.

Real-time notifications for performance anomalies.

Recruit-ready player summaries for scouting and external evaluations.

Impact

The Player Development & Profile System turns our data into a competitive advantage:

Coaches can make data-driven decisions about player development and strategy.

Players gain clear visibility into their progress, motivating them with tangible goals.

The team has a centralized, scalable platform that grows as new data is collected each season.

This project bridges the gap between traditional coaching methods and modern data analytics, bringing professional-grade tools to our university baseball program.

As a research assistant in the Lipscomb University Robotics Lab, I supported a project focused on training robots to make autonomous decisions using Deep Reinforcement Learning (DRL). While I did not lead the project, I contributed by preparing datasets, cleaning experimental data, and visualizing test results, ensuring the research team had reliable data for modeling and analysis.

This experience gave me hands-on exposure to DRL workflows, helping me understand how algorithms like Markov Decision Processes (MDPs) and unsupervised learning can be applied to robotics.

Project Goals

Develop models to teach robots to perform complex actions, such as navigation, object manipulation, and task selection.

Use DRL algorithms to simulate trial-and-error learning, allowing robots to optimize their decision-making over time.

Build tools to analyze and visualize robot performance, making research results interpretable for the team.

My Role & Contributions
1. Data Preparation and Cleaning

Processed large volumes of experimental data collected from robot tests and simulations.

Normalized and structured data for use in DRL training pipelines, ensuring consistency and accuracy.

Removed corrupted or incomplete logs to prevent model degradation.

2. Data Visualization

Created clear graphs and charts to track:

Reward progression over episodes.

Policy performance across different training iterations.

Comparison of test outcomes between models.

Visualizations helped the team evaluate experiments quickly and identify promising approaches.

3. Collaboration & Support

Worked alongside graduate students and researchers, assisting during robot trials and debugging sessions.

Helped bridge the gap between raw data and actionable insights, supporting decisions on model adjustments and testing strategies.

Fake News Detection – NLP Classification System

The Fake News Detection Model is a machine learning project designed to classify news articles as either Real or Fake, providing a confidence score to reflect certainty. Developed as part of a 7-person cross-functional team during my time with Accenture AI Studio, this project demonstrates my ability to collaborate effectively, manage complex workflows, and apply advanced NLP (Natural Language Processing) and machine learning techniques to real-world problems.

I contributed to both technical development and project management, participating in every stage of the project lifecycle, from initial planning and data preparation to model deployment and final presentation.

Project Objectives

• Build a binary classification system to detect fake news articles with a high degree of accuracy.
• Create a transparent and explainable solution that users and stakeholders can trust.
• Develop a scalable, API-driven backend for integrating the model into other platforms.
• Ensure smooth collaboration and clear communication across all team members and tasks.

My Role & Contributions
Project Management

• Coordinated team activities by setting milestones, managing timelines, and ensuring alignment between data, modeling, and deployment teams.
• Led regular stand-up meetings, facilitating updates and resolving blockers quickly.
• Oversaw documentation and reporting, ensuring deliverables were clear, accurate, and professional.
• Acted as a bridge between technical contributors and non-technical stakeholders, translating technical progress into actionable insights.
  
  Technical Contributions
  

Participated hands-on in each development phase, including:

• Data Preparation – Cleaning and preprocessing text data (HTML cleanup, stop-word removal, lemmatization, and normalization).
• Model Building – Implementing baseline algorithms such as Logistic Regression and Support Vector Machines (SVM) with TF-IDF features.
• Advanced Modeling – Fine-tuning DistilBERT, a transformer-based model, for improved accuracy and contextual understanding.
• Evaluation & Validation – Designing evaluation strategies using metrics like F1-score, ROC-AUC, and Precision/Recall curves to measure performance and mitigate class imbalance.
• Deployment – Assisting in building a Flask-based API endpoint to serve predictions for external applications.

By combining project management and technical work, I ensured that our team remained organized and focused, while also contributing directly to the codebase and model development.