From data to models. From models to applications.

RiverSort AI

A computer vision prototype that detects floating plastic waste in river images, tracks objects across video frames, and provides transparent, rule-based cleanup recommendations.

Held-out FPWR test (440 images, 4 classes): mAP50 0.6936 · macro-F1 0.6904 at confidence 0.40 and IoU 0.50. Dataset-specific results; performance may differ in other river conditions.

Credit Risk Prediction

A machine learning workflow for credit default risk modeling, combining feature engineering, model evaluation, and profitability simulations to explore lending strategies.

Pleiades Star Cluster Membership Analysis

A modular HDBSCAN pipeline for identifying candidate Pleiades members in Gaia data, with preprocessing, reference-catalog comparisons, visualization, and automated tests.

BCA Mobile App Sentiment Analysis

An NLP workflow for analyzing BCA Mobile user reviews, with Indonesian text preprocessing and machine learning models for sentiment classification.

Smart Cancer AI

An educational application integrating three CNN-based models for brain, breast, and lung image classification. A demonstration of AI in a web interface, not a clinical diagnostic tool.

Book Recommendation System

A content-based recommender that uses TF-IDF and cosine similarity to suggest books with metadata similar to an input title.

Solana Price Forecasting

A time-series forecasting experiment using LSTM to explore historical Solana price patterns.