A Senior AI Engineer with a rare dual expertise: architecting low-level Edge AI systems in C/C++ and NVIDIA GPU optimization, while designing intelligent multi-agent autonomous workflows with LLMs and Generative AI.
I take ideas from algorithm design all the way to hardware realization. Driven by first-principles engineering, I thrive on solving bottlenecks in edge environments — making sure AI models don't just perform well on paper, but execute flawlessly under strict latency and memory constraints in the real world.
Right now I'm building intelligent autonomous systems that remove operational friction for real people — from Indonesian SMEs running 24/7 autonomous business agents, to enterprise teams automating fraud detection and risk analytics. Over 10+ years I've delivered sub-66.7 ms inference across 15 simultaneous 1080p streams on one end, and enterprise agentic systems on the other.
Alongside building, I advise and teach — currently as an AI Architect & Technical Advisor at V-TEKI, a technical mentor at Garuda Hacks and Google's developer programs, and for over a decade as a lecturer and researcher in Electrical Engineering at Soegijapranata Catholic University, where I've led national robotics teams across UAV, AUV, AGV, and humanoid platforms.
TensorRT (FP16/INT8 quantization), NVIDIA Jetson, PyTorch Forward Hooks, dynamic batching, and strict latency budgeting for real-time inference.
QwenPaw (AgentScope), Qwen3-14B, LangChain, LangGraph, MCP, and the Claude & Gemini APIs — building autonomous, tool-using agent workflows.
Computer vision, CNNs, Transformers, sensor fusion, and the YOLO family for detection, recognition, and verification.
Tesseract OCR and Indonesian KTP extraction with regex-based structured field parsing and image quality gating.
ROS, UAV / AUV / AGV systems, and microcontroller integration across Arduino, STM32, Raspberry Pi, and dsPIC.
Domain-Driven Design, async FastAPI, Go microservices, Docker, and GitHub Actions CI/CD.
ETL pipelines, RAG-based knowledge retrieval, and data visualization with Pandas, Plotly, and Tableau.
Continued learning — selected 2025–26 programs & bootcamps:
Google Scholar — 40 citations · h‑index 4 · Scopus — 18 citations · h‑index 3