I got into AI when I saw how fast it was changing everything around me and arning became my obsession. That curiosity turned into a real drive to build systems that actually work, not just models that live in notebooks, So I put that obsession to work.
I'm an Informatics Engineering graduate (GPA 3.79) who also completed Bangkit Academy, a 6 month intensive program led by Google, GoTo & Tokopedia. I spent those years building and deploying real AI projects, from an ecommerce customer service agent chatbot to a credit scoring app, with every project shipped live and ready to try right now.
I enjoy the whole process, from messy data to something that actually helps people, and I'm still learning every day, which is exactly what keeps me going.
A production grade RAG system that answers HR policy questions using the GitLab Handbook as its knowledge base. Features hybrid search (Vector + BM25 + RRF), bilingual support, streaming responses, conversation memory, and a full monitoring stack with Prometheus and Grafana.
An agentic RAG system for e-commerce customer service with 5 specialized agents for FAQ, Product, Order, Promo, and Escalation each with their own knowledge base. Features intent classification, hybrid retrieval (BGE-M3 + BM25 + RRF), and LangSmith tracing, deployed on Hugging Face Spaces.
An end-to-end Probability of Default model trained on Home Credit data. Features rigorous validation including Out-of-Time testing, PSI stability check, fairness analysis by age group, and cutoff policy simulation with stress testing and deployed as an interactive Streamlit app.
A fraud detection model trained on the IEEE-CIS dataset with severe class imbalance. Features custom threshold tuning optimized for cost savings (67% estimated reduction), a full preprocessing pipeline, and deployed with drift monitoring baseline on Hugging Face Spaces.
A deep learning model for kidney stone detection from ultrasound images using Swin Transformer, achieving 99.57% accuracy and outperforming CNN and Vision Transformer (ViT) baselines. Trained on 9,396 images and published in JATI EMAS journal (Vol. 9, 2025), with a live demo deployed on Hugging Face Spaces.
My core focus is building and deploying arning systems from data preprocessing and model training to real-world deployment. I work mostly in Python and enjoy the full pipeline, from raw data to a working product.
I'm currently open to new opportunities, whether it's a full time role or just a conversation about AI and machine learning