A roadmap and tracker for mastering Applied AI Engineering. Focuses on moving from "API Wrapper" to Model Builder. Covers Fine-tuning, Evaluation, and Local Inference. Features polyglot tips for PHP/Java/Node/C# devs bridging the gap to Python.

No description available

A machine learning project that predicts the likelihood of road accidents based on factors like weather, time, location, and vehicle type. It uses data preprocessing, feature engineering, and ML models to identify high-risk conditions and improve road safety insights.

A complete data engineering and ML pipeline for traffic anomaly detection using Graph Neural Networks. Features a reproducible research workflow, ablation studies, and an interactive HTML dashboard to explore detected anomalies across NYC road segments.

A Python project to analyze and predict the severity of road accidents in the UK using machine learning. This project includes data preprocessing, feature engineering, categorical encoding, and a framework for training and evaluating ML models.

A simple yet insightful Machine Learning project to predict housing prices based on multiple factors such as house area, number of bedrooms, furnishing status, and proximity to main roads. This project demonstrates end-to-end ML workflow including data preprocessing, feature engineering, model building, and evaluation.

Built an end-to-end ML system for traffic intelligence using real-world sensor data. The project forecasts short-term traffic speed, classifies road conditions (Jam/Slow/Free), and discovers hidden traffic patterns with K-Means. Focused on time-series feature engineering, time-aware validation, robust ML pipelines, and ready model deployment.

A machine learning regression project completed for Boğaziçi University’s Math482 course, based on the Kaggle competition Predicting Road Accident Risk. The project covers the full ML pipeline — from EDA, preprocessing, and feature engineering to model training and evaluation — achieving strong results with LightGBM.

Analyzed UK road accident data to predict severity using ML models like Logistic Regression, Random Forest, XGBoost, and tuned LightGBM. Applied EDA, feature engineering, SMOTE, class weighting, and hyperparameter tuning. Used DBSCAN to identify accident hotspots. Best model achieved macro F1 of 0.42 and 19.3% recall for fatal accidents.