I build production-grade AI systems for real government workloads. RAG agents, face recognition at scale, Rust backends — serving 15,000+ civil servants daily inside Pemerintah Kabupaten Sumenep. Currently pursuing S2 at ISTTS Surabaya, thesis on evaluating RAG for Indonesian government regulation.
I build AI systems that actually work — not demos, but production systems handling thousands of real users every day inside the Indonesian government.
My work sits at an unusual intersection: deep AI engineering combined with government domain expertise. I understand both how to implement ArcFace with FAISS at scale, and how Indonesian civil service regulation (ASN law) is structured — which makes building systems like iASN possible.
Currently leading the full AI and digital infrastructure at BKPSDM Kabupaten Sumenep, where I've rebuilt the entire civil service management ecosystem — Rust backends, Next.js frontends, Flutter mobile, and an AI layer powered by Gemini and Qdrant.
My S2 thesis focuses on evaluating RAG-based AI agents for Indonesian government employment regulation — using RAGAS metrics and custom evaluation frameworks for low-resource Bahasa Indonesia legal documents.
An advanced AI agent evolved from simple RAG into a full orchestrator with native tool calling (Gemini SDK). Answers complex queries about civil service regulations (UU No. 5/2014), retrieves real-time employee data from SIMPEG/SIAGA, and renders interactive data visualizations. Features Parent-Child Retrieval, Hybrid Search with RRF, Semantic Caching, Vision LLM document assessment, and integrated RAGAS evaluation.
Glassmorphism-styled Single Sign-On portal for all civil service apps. MFA with OTP biometric, dynamic RBAC, Web-OS aesthetic with animated transitions.
Production task scheduler and secure proxy. Rust control plane, Docker sandboxing, AES-256-GCM secret vault, Gemini AI assistant for script generation and debugging.
Tri-Mode AI for civil service data requests: SQL, Code, Hybrid. Gemini-powered narrative insights with PII protection and sandboxed Python execution via Pandas.
Full rebuild from PHP/jQuery to modern stack. Decoupled Next.js + Actix-Web, multi-PostgreSQL, TPP calculation automation, glassmorphic premium UI.
Daily performance management for ASN. Rebuilt from legacy, integrated with SIMPEG & SIMANTRA ecosystem, real-time validation, GCS document storage.
Face recognition attendance with geofencing (30m radius), ML Kit detection, liveness validation. Migrating VGGFace → ArcFace + FAISS IVF for 15k employees.
Core HR management API. BKN SIASN real-time sync, OpenCV biometric liveness, PDF/Excel/Word document generation engine, automated cron data archiving.
Offline-first POS for Indonesian SMEs. Clean architecture, SQLite + Firestore sync, debt management, movement-based inventory. Flutter + Riverpod.
Comprehensive evaluation of RAG chatbot (iASN) for ASN queries on UU No. 5/2014. Custom metrics for low-resource Bahasa Indonesia legal domain. Ablation study across retrieval components.
Comparative study: RAG-only vs RAG + LoRA fine-tuned model for ASN regulation. Synthetic dataset from regulatory documents via self-instruct. Target model: Llama 3 / Qwen2.5.
Migration from VGGFace + Qdrant to ArcFace with FAISS IVF indexing. MiniFASNet passive liveness + MediaPipe active challenge for anti-spoofing in civil servant attendance system.
First ASR system for Bahasa Madura, spoken by ~7 million people in East Java. Minimal digital resources. Potential for novel dataset creation and transfer learning from Bahasa Indonesia models.
"Open to research collaboration, AI consulting, and building technology that creates real impact in the public sector."
Available for AI research collaboration, technical consulting on government AI systems, and academic partnerships. Based in Sumenep, East Java — remote-friendly.