model-evaluation

Fine-tune anime-seg (ISNetDIS) bằng LoRA trên 1306 cặp ảnh-mask để tách nền anime; có split cố định, evaluation trên mask thô, hậu xử lý alpha_gain cho RGBA cutout, merge adapter vào checkpoint, và CLI rmbg cho train/eval/predict/report/make-split với manifest tái lập đầy đủ.

Classification evaluation, calibration, slice analysis, and drift reports with a dependency-light Python CLI

A demonstration of MLflow working end-to-end on sample data sets.

A data science project made to develop a predictive model for estimating house prices in King County,

This project based on car price prediction using a Random Forest model and a Streamlit web app for a user-friendly interface.

AutoMLPrediction is a comprehensive, automated machine learning (AutoML) platform designed to simplify the process of building, training, and deploying machine learning models. The platform leverages state-of-the-art AutoML techniques to enable users, regardless of their technical expertise, to make accurate predictions from data effortlessly.

Student Performance Prediction This project examines the student performance data, pre-processes the features, and implements machine learning methods (Random Forest) for the forecasting of final grades. The data is subjected to exploratory data analysis (EDA) and feature engineering prior to model training and assessment.

End-to-end MLOps pipeline for hotel booking demand forecasting. Includes modular components for data ingestion, model training, evaluation, versioning, and deployment. Features configuration-based execution, CI/CD with GitHub Actions, and automated logging and testing.

A Vision Transformer (ViT) implementation for image classification using CIFAR-10 dataset, leveraging HuggingFace's Trainer API for computational efficiency

A comprehensive pipeline for generating, analyzing, and evaluating models using AI as Judge model

End‑to‑end ML project predicting loan approvals using 9 applicant features. Data is processed, modeled with Random Forest and GridSearchCV, deployed via FastAPI and Uvicorn, with a simple frontend, containerized in Docker, tested with pytest, and automated through CI/CD.

Machine learning project for classifying cybersecurity incidents (TP, BP, FP) using the GUIDE dataset. Includes data preprocessing, feature engineering, model benchmarking, and evaluation with macro-F1, precision, and recall. Supports SOC automation, threat detection, and enterprise security management.

WinQ leverages machine learning to predict wine quality using key physicochemical features, delivering actionable insights with strong model accuracy. Developed for Stanford Code in Place 2025, this project showcases the power of Python and data science fundamentals in a real-world context.

End-to-End NLP project that analyzes mobile customer reviews using TF-IDF and Logistic Regression to deliver actionable business insights with ~87% accuracy. Demonstrates real-world machine learning workflow from EDA to model deployment.

10Pearls Shine Internship, End-to-end ML pipeline predicting Karachi AQI. Fully automated via GitHub Actions. Watch video at: https://youtu.be/nTy6kUEoIws

Learning representations on MNIST: Softmax → MLP → CNN → Augmentation, with metrics, visualizations, and error analysis.

Machine Learning models to predict Netflix user churn (MSBA group project).

Comprehensive machine learning repository showcasing feature engineering, hyperparameter optimization, and rigorous model evaluation metrics for certification validation.