Muhammad Haerul's on Medium
An in-depth engineering guide on designing low-latency multimodal agent workflows, edge computer vision pipelines with YOLO, and resilient full-stack AI deployment architectures.
AI Full-Stack Specialist
My work sits at the intersection of AI engineering, applied research, product development, and education technology. I develop end-to-end AI solutions, from data preparation and model development to application integration and deployment, while exploring new approaches to make AI more practical, reliable, and impactful. Beyond engineering, I help individuals and organizations understand, adopt, and apply AI through training, mentorship, and hands-on learning.
I enjoy bridging the gap between cutting-edge AI research and real-world execution, translating breakthrough models into scalable products that solve concrete operational challenges. My cross-disciplinary experience spans high-growth tech industry, public-sector digital transformation, and nation-wide AI education initiatives, with a relentless focus on reliable software design and ethical, human-centered technology.
My mission is to build, research, and teach AI solutions that create compounding real-world value — shipping intelligent systems and dependable software architectures that make artificial intelligence truly practical, performant, and empowering for people, businesses, and society.
Universitas Hasanuddin
September 2020 — July 2024
AksaraHub Nusantara · Hybrid · Makassar, Indonesia
Leading AI architecture, product development, and data analytics for AksaraHub.id — Indonesia's premier cultural-tech initiative dedicated to the digital preservation and revitalization of regional scripts and languages (Bugis, Makassar, Mandar, and Toraja). Architecting end-to-end computer vision and deep learning OCR pipelines for real-time handwritten script recognition, neural machine translation, and interactive digital canvas tools.
RevoU · Remote
Mentoring and upskilling hundreds of professionals and university students across foundational through advanced AI domains — specializing in AI fundamentals, prompt and context engineering, agentic workflows, and production Generative AI applications. Designed and delivered tailored enterprise and public curricula for government bodies, corporate leaders, and academic institutions.
Paper.id · AI Product & Enablement, Jakarta
Led development of an LLM-powered OCR system for KYC/KYB processing (99% accuracy, 4s per document), a multi-agent HR chatbot (99% response accuracy across 1,000+ queries), and a hybrid keyword + vector + LLM search engine. Built the ImageProcessingOCR enhancement pipeline that lifted OCR accuracy by 10% while cutting file size 40% and processing time 50%.
Masa Geosciences · Remote, Jakarta
Researched, developed, and deployed computer vision models and machine learning pipelines for automated geological mineral identification, color segmentation, and mineral content estimation.
XL Axiata · On-site · Jakarta, Indonesia
Built four production computer-vision systems for the X-Camp incubation program: a road-damage detector across 45,000+ images (78% accuracy), a PPE detector for industrial safety running on Raspberry Pi (84% accuracy), a headcount monitoring dashboard, and a TensorFlow-based fuse anomaly detector.
Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi · On-site · Jakarta
I was part of Artificial Intelligence RnD Project focusing on Developing Education Technology.
Scientific papers, research publications, and open-source technical documentation.
An in-depth engineering guide on designing low-latency multimodal agent workflows, edge computer vision pipelines with YOLO, and resilient full-stack AI deployment architectures.
Galaxy Object Detection Core is a modular, production-ready object detection library built on YOLO and YOLOWorld architectures. Designed for real-time performance and high-precision detection, the package supports 20+ industry categories including safety gear, documents, vehicles, road damage, KTP, and minerals.
Undergraduate scientific research and university repository publication analyzing, developing, and benchmarking computer vision models for Personal Protective Equipment (PPE) compliance detection in industrial environments.
Radio Republik Indonesia (RRI Mamuju) & PCMI Sulawesi Barat
Studio RRI Pro 2 Mamuju (Live Broadcast 89.3 FM & YouTube)
Yayasan Khouw Kalbe
Zoom Meeting
Google Student Ambassador
Makassar Government Center
Google Developer Group on Campus Makassar
Universitas Ciputra Makassar
Universitas Sulawesi Barat
Hotel Villa Bogor, Majene
Badan Siber dan Sandi Negara (BSSN)
Zoom Meeting
Authentic ratings and impressions submitted by participants across AI workshops and training sessions.
Written recommendations from colleagues, engineering leaders, and project collaborators.
Collaboration inquiries, technical speaking invitations, and project consultations.