Hi, I'm Arizal Firdaus

Data Scientist | ML & DL Specialist

Transforming Data into Intelligence

Passionate about extracting meaningful insights from complex datasets using advanced machine learning and deep learning techniques. Specialized in building intelligent systems that drive business value and innovation.

View My Work

About Me

A results-driven Data Scientist with a deep specialization in Machine Learning and Deep Learning models. I transform complex datasets into high-performance, intelligent systems that drive business growth, enhance efficiency, and uncover hidden opportunities.

My passion lies in exploring the entire data lifecycle: from rigorous data analysis and feature engineering to building, training, and fine-tuning complex neural networks. I am driven by a practical, hands-on approach to ensure every model is not just technically sound, but also delivers measurable value.

Featured Projects

Kaggle Competition Top 9% Rank

This Kaggle Playground Series (S5E7) project predicted personality type (Introvert vs Extrovert) from survey responses. Achieved Top 9% (350/4,329) and received a Bronze Medal for notebook quality and reproducibility. Utilized advanced feature engineering and hyperparameter tuning with Gradient Boosting to enhance model performance.

Python Scikit-learn Pandas Hyperparameter Tuning Gradient Boosting

View Detail View Notebook

Deep Learning for Stock Price Forecasting

I built a complete end-to-end pipeline for this time-series project. The process included cleaning and preprocessing a 10-year raw dataset from Investing.com. I then scaled the data and transformed it into supervised learning 'windows' to feed into a stacked Long Short-Term Memory (LSTM) neural network built with TensorFlow/Keras. Finally, I deployed the trained model into an interactive web application that provides a 7-day forecast.

TensorFlow Keras LSTM Time Series Forecasting Seaborn

View Detail Live Demo

AI Object Tracking & Counting for Traffic Analysis

Developed a proof-of-concept for automated traffic analysis using Python, PyTorch, and the SORT algorithm. Leveraging Facebook's DETR model, the system provides a cumulative count of unique objects as they pass through a designated 'counting zone'.This baseline successfully demonstrates the core tracking methodology. While functional, accuracy can be refined for complex object occlusions, making it a strong foundation for more advanced, production-ready analytics solutions.

Object Tracking Computer Vision PyTorch OpenCV ResNet50