Legal Document Intelligence System
Analyzes Indonesian legal documents and regulations and produces structured daily briefs that inform policy responses at Komdigi, Indonesia's Ministry of Communication and Digital Affairs.
Machine Learning Engineer
Muhammad Zidan Arsyad
I build and ship end-to-end LLM systems — retrieval, fine-tuning, and production serving.
- RAG systems
- LLM fine-tuning (SFT/CPT)
- vLLM serving
About
I am an ML engineer with hands-on experience building and deploying end-to-end LLM pipelines — RAG, SFT, and CPT — for production use in government intelligence workflows. Most recently I led a four-person AI team delivering a legal document intelligence system for Komdigi (Indonesia's Ministry of Communication and Digital Affairs), serving structured daily briefs to ministry officials. I work across the full ML lifecycle: data collection, model training, API orchestration, and cloud deployment.
Portfolio
Selected work
Airline Review NLP Pipeline
Scraped 1,000+ Skytrax customer reviews and built an end-to-end NLP pipeline — sentiment analysis with NLTK and topic modeling with Gensim LDA — surfacing the key drivers of customer satisfaction for stakeholder review.
Air Quality & Volatility Forecasting
Developed ARMA models to forecast PM2.5 air-quality trends in Jakarta, and built a FastAPI service for real-time stock volatility prediction on Alpha Vantage market data.
Damage & Bankruptcy Prediction
Predicted earthquake building-damage levels in Nepal with logistic regression and decision trees, and built bankruptcy prediction models on heavily imbalanced datasets using Random Forest and Gradient Boosting.
Education
2026 — present
Master of Computer Science — Artificial Intelligence
Gadjah Mada University
Activity: Artificial Intelligence Talent Factory (AITF) UGM.
2018 — 2022
Bachelor of Engineering — Computer Engineering
University of Indonesia
GPA 3.71/4.00 · Activity: IME FTUI, Jogjakun UI.
Résumé
Contact
Let's build something.
Open to ML engineering and applied-research work. The fastest way to reach me is email.