Highlighted Projects

Retrieval-Augmented Chatbot for Document-Based Q&A

This project is an OpenAI-powered chatbot that uses Retrieval-Augmented Generation (RAG) to enhance responses with relevant information from provided documents.

Key Features
• Deployed using Docker and Kubernetes for scalable service delivery;
• Accessible through a user-friendly web interface;
• Utilizes OpenAI's GPT-3.5 model for natural language processing;
• Built with LangChain to orchestrate LLM workflows efficiently;
• Implements Retrieval-Augmented Generation (RAG) for context-aware responses;
• Features document retrieval capabilities for context-enhanced responses.

Fraud Detection Graph Visualization with D3.js

This project uses Graph Neural Networks (GNN) to enhance fraud detection in online gambling and financial transactions in Indonesia. By analyzing transaction networks, it uncovers hidden patterns and detects fraud syndicates more effectively than traditional rule-based methods.

Key Features
• Large-scale graph processing for complex transaction networks;
• Real-time fraud detection capabilities;
• Interactive network visualization tools;
• Node embeddings and K-means clustering for pattern recognition;
• Syndicate scoring system to identify fraudulent user groups.

Youtube Video Recommendation via NLP Keyword Extraction

One of my favorite side projects is a YouTube video recommendation insights tool, where I used the YouTube Data API and NLP to analyze watch history and uncover recommendation patterns.

Key Features
• OAuth 2.0 integration for secure user authentication;
• Transcript analysis and keyword extraction to identify trends;
• Visualization features for better data interpretation;
• Data export capabilities for further analysis.

Customer Goods Data Modeling, Clustering, and Segmentation

Another side project is a Customer Goods Data Modeling project that tackles industrial challenges through predictive modeling for daily sales quantity and customer segmentation.

Key Features
• Utilized PostgreSQL and DBeaver for efficient data ingestion;
• Built interactive dashboards using Tableau Public;
• Developed predictive models in Python using Google Colab;
• Applied time series ARIMA for sales forecasting;
• Used clustering techniques for customer segmentation.

• Downloadable Portfolio

Achievements & Publications

PUBLISHED: Propagation Characteristics of Madden Julian Oscillation in the Indonesian Maritime Continent: Case Studies for 2020-2022, Agromet Journal, doi: 10.29244/j.agromet.38.1.1-12PREPRINT / UNDER REVIEW: An Elementary Approach to Predicting Indonesian Monsoon Index: Combining Ann-Arima Hybrid Method and Practical Use

• ResearchGate

Professional Certifications

Coursework Certifications