
Build Face Search using PostgreSQL Vector
Face recognition technology has become a key component of many applications, from security systems and mobile device a...
Computer Vision & Edge AI Engineer
Computer Vision Engineer with 5+ years of professional experience designing, developing, optimizing, and deploying production-grade Computer Vision and Edge AI systems for industrial automation.
Specialized in Object Detection, Semantic Segmentation, OCR, Object Tracking, Multi-Camera Vision, and AI-based Inspection Systems, with strong expertise in taking AI solutions from dataset development and model training to production deployment and system integration.
Experienced in deploying real-time AI workloads on NVIDIA Jetson edge platforms using TensorRT, ONNX Runtime, and DeepStream, with a strong focus on inference optimization, latency, reliability, and resource efficiency.
Proven experience architecting and delivering automated container inspection systems for large-scale port terminal operations, integrating AI vision, multi-camera systems, LiDAR, Industrial IoT, PLC, and backend services into production environments.
Strong background in AI system architecture, edge computing, microservices, industrial automation, and production Computer Vision, with hands-on experience delivering systems designed for high-throughput and real-time operational environments.
Semi Automation labeling for OCR dataset
Fine tuning opensource model OCR
Implement object detection and object classification for real world case OCR Plate Indonesian
Build yoga pose estimation with pose point classification
Face search using PostgreSQL vector embeddings
Low-latency Python streaming, single-image capture, and ISAPI encoding controls for Hikvision cameras
Build the Linux kernel and modules on board the Jetson AGX Orin, Orin Nano or Orin NX
Python library and CLI for reading UHF RAIN RFID (EPC Gen2) tags from Invengo fixed readers over TCP
Real-time container inspection for port terminal operations. LiDAR-triggered multi-camera and line-scan pipelines read Container Number, ISO Code, and Max Gross via OCR, verify seal presence, and detect rust and damage on every side of the container — deployed on NVIDIA Jetson edge devices.
Gantry crane OCR portal that captures containers from 5 camera views — left, right, rear, front, and top — during crane operations. Recognizes ISO 6346 container numbers with check-digit validation, ISO size/type, and CSC plate max gross, flags seal presence status, and logs every transaction to a real-time reports dashboard with sub-second processing (200–450 ms).
Image processing pipeline that stitches frames from a moving top-view camera into a single full-length panorama of the container as the truck passes under the gantry — turning standard video into complete top-side inspection imagery, with frame registration and seam blending across the stitched output. One stitched image is produced per transaction for the OCR portal.
Live camera scanner for halal product verification, powering the Scan Aman Halal app. A captured frame of the packaging is embedded and matched against a product database through vector similarity search — instantly returning the product and its halal certificate status (name and producer). Three lookup modes: product name, packaging scan, and composition check.
Real-time PPE (Personal Protective Equipment) violation detection across multiple cameras in industrial container yards. The system detects persons and flags violations such as NON_HELMET and NON_VEST with severity levels, then logs every event with camera location, timestamp, and a cropped evidence snapshot into a review dashboard — where operators can add review notes and mark records reviewed, dismissed, or deleted.
Custom vehicle classification for Indonesian toll roads (v1.0 Jasamarga) — classifies passing vehicles into the official toll classes (Golongan 1A, 1B, 2A, 2, 3) from CCTV footage in real time, deployed on NVIDIA Jetson series edge devices.
Real-time face recognition attendance system running on live CCTV — recognizes registered personnel walking through building entrances and records their attendance automatically, with identity and confidence overlay drawn on the video feed at the gate.

Face recognition technology has become a key component of many applications, from security systems and mobile device a...

LLaMA OCR stands for “Large Language Model Application Optical Character Recognition”. It’s an optical character recog...

What is Oriented Bounding Boxes Object Detection? Continue reading on Medium »...
"Panorama Reconstruction Using Farneback Optical Flow and Kalman Filter for Moving Object Mapping."
Reach out for industrial vision audits, edge inference architecture on Jetson, or custom model optimization:
Computer Vision & Edge AI Engineer
Computer Vision Engineer with 5+ years of professional experience designing, developing, optimizing, and deploying production-grade Computer Vision and Edge AI systems for industrial automation.
Specialized in Object Detection, Semantic Segmentation, OCR, Object Tracking, Multi-Camera Vision, and AI-based Inspection Systems, with strong expertise in taking AI solutions from dataset development and model training to production deployment and system integration.
Experienced in deploying real-time AI workloads on NVIDIA Jetson edge platforms using TensorRT, ONNX Runtime, and DeepStream, with a strong focus on inference optimization, latency, reliability, and resource efficiency.
Proven experience architecting and delivering automated container inspection systems for large-scale port terminal operations, integrating AI vision, multi-camera systems, LiDAR, Industrial IoT, PLC, and backend services into production environments.
Strong background in AI system architecture, edge computing, microservices, industrial automation, and production Computer Vision, with hands-on experience delivering systems designed for high-throughput and real-time operational environments.
Semi Automation labeling for OCR dataset
Fine tuning opensource model OCR
Implement object detection and object classification for real world case OCR Plate Indonesian
Build yoga pose estimation with pose point classification
Face search using PostgreSQL vector embeddings
Low-latency Python streaming, single-image capture, and ISAPI encoding controls for Hikvision cameras
Build the Linux kernel and modules on board the Jetson AGX Orin, Orin Nano or Orin NX
Python library and CLI for reading UHF RAIN RFID (EPC Gen2) tags from Invengo fixed readers over TCP
Real-time container inspection for port terminal operations. LiDAR-triggered multi-camera and line-scan pipelines read Container Number, ISO Code, and Max Gross via OCR, verify seal presence, and detect rust and damage on every side of the container — deployed on NVIDIA Jetson edge devices.
Gantry crane OCR portal that captures containers from 5 camera views — left, right, rear, front, and top — during crane operations. Recognizes ISO 6346 container numbers with check-digit validation, ISO size/type, and CSC plate max gross, flags seal presence status, and logs every transaction to a real-time reports dashboard with sub-second processing (200–450 ms).
Image processing pipeline that stitches frames from a moving top-view camera into a single full-length panorama of the container as the truck passes under the gantry — turning standard video into complete top-side inspection imagery, with frame registration and seam blending across the stitched output. One stitched image is produced per transaction for the OCR portal.
Live camera scanner for halal product verification, powering the Scan Aman Halal app. A captured frame of the packaging is embedded and matched against a product database through vector similarity search — instantly returning the product and its halal certificate status (name and producer). Three lookup modes: product name, packaging scan, and composition check.
Real-time PPE (Personal Protective Equipment) violation detection across multiple cameras in industrial container yards. The system detects persons and flags violations such as NON_HELMET and NON_VEST with severity levels, then logs every event with camera location, timestamp, and a cropped evidence snapshot into a review dashboard — where operators can add review notes and mark records reviewed, dismissed, or deleted.
Custom vehicle classification for Indonesian toll roads (v1.0 Jasamarga) — classifies passing vehicles into the official toll classes (Golongan 1A, 1B, 2A, 2, 3) from CCTV footage in real time, deployed on NVIDIA Jetson series edge devices.
Real-time face recognition attendance system running on live CCTV — recognizes registered personnel walking through building entrances and records their attendance automatically, with identity and confidence overlay drawn on the video feed at the gate.
Face recognition technology has become a key component of many applications, from security systems and mobile device authentication to personalized experiences...
Read on Medium arrow_forward
LLaMA OCR stands for 'Large Language Model Application Optical Character Recognition'. It's an optical character recognition pipeline leveraging multimodal LLMs...
Read on Medium arrow_forward
What is Oriented Bounding Boxes Object Detection? Oriented bounding boxes allow detecting objects at varying rotation angles for industrial vision...
Read on Medium arrow_forwardBachelor of Informatics Engineering • GPA: 3.3 / 4.00
Undergraduate Thesis: Panorama Reconstruction Using Farneback Optical Flow and Kalman Filter for Moving Object Mapping
Software Engineering (Vocational High School)
Need an on-premise vision POC, Jetson deployment optimization, or industrial OCR pipeline? Send a project brief.