- Product Matching | Computer Vision + NLP | Kaggle Notebook 
 Determine if two e-commerce products are the same by their images.
 Silver medal - Top 5% (101 from 2,426) in Kaggle’s Featured Code Competition - Shopee Price Match Guarantee.
 Tech stack: PyTorch, Numpy, TensorFlow
 Example:  
 Architecture:  
- Information Extraction | NLP | Source Code 
 Extract part of interest (POI) and street names from Indonesian address.
 Top 3% (28 from 1,034) in Shopee Code League Data Science Challenge 2021.
 Tech stack: PyTorch, Numpy
 Example:- raw address - Extacted POI - Extracted street - karang mulia bengkel mandiri motor raya bosnik 21 blak kota - bengkel mandiri motor - raya bosnik - primkob pabri adiwerna - primkob pabri - <null> - jalan mh thamrin, sei rengas i kel. medan kota - <null> - jalan mh thamrin - Architecture:   
- e-Commerce Product Attributes Extraction | NLP | Source Code 
 Extract multi-attributes from the product titles in the Indonesian e-commerce platform.
 It’s a sequence labelling task using BERT for Token Classification model. I employ this work to preprocess the product titles on the Shopee Product Matching challenge, which led to my first Kaggle’s silver medal.
 Tech stack: PyTorch, Numpy
 Example: - Architecture:   
- Photovoltaic Failure Detection | Computer Vision | Short Documentation 
 Automatic Failure Detection in Photovoltaic Solar Panel Systems
 Tech stack: PyTorch, OpenCV  
- Toxicity Detection (Content Moderation) | NLP | Source Code - Documentation - Slide 
 A content moderation solution for a social media platform to highlight toxic content, i.e., any content that could be racist, derogatory, abusive or demeaning towards other members on the platform.
 Tech stack: PyTorch, Numpy  
- PiCirclet | Computer Vision + TTS | Video - Attachment 
 A device for visually impaired people to help them hear what they see. It’s powered by a RaspberryPI as a processing unit. It captures an image using a camera, recognizes texts (optical character recognition) and then outputs voices (speech synthesize)
 1st Winner at a Competition on Building Device for Visually Impaired People held by Syamsi Dhuha Foundation  
- Robot for Soccer | Computer Vision | Video1 - Video2 
 Computer vision: (1) Implement robot’s localization by combining visual odometry and landmark detection; (2) Implement object detection algorithm, i.e. for detecting a ball, opponent robots, and a goal-post.
 Tech stack: Python, C++, OpenCV, ROS(Robot Operating System)  
- Object Detection | Computer Vision 
 Develop an bject detection using Faster RCNN. This project is a minor work during internship at Bukalapak as a Machine Learning Research Engineer
 Tech stack: Python, TensorFlow  
- Alinamed | Software Developmet | Android - iOS - Web 
 A medical apps that connect doctor with patient through online booking.
 Tech stack: Backend (PHP, Golang), Web Frontend (HTML, CSS), App Frontend (React Native), Infra (AWS)   
- Citra | Computer Vision | Source Code - Online Demo 
 A web app to perform simple image processing techniques, i.e., Histogram equalization, Histogram match (specification), Character detection using chain code, and Thinning using Zhang Suen. These algorithms are fully coded from sratch using Javascript. This is an undergraduate course project for IF4073 Image Processing
 Tech stack: Javascript   
- Alumni Salman | Software Developmet 
 An Android application for Masjid Salman ITB. My friends and me did all the software development process from requirement gathering until product delivery. This is an undergraduate course project for IF3250 Software Project
 Tech stack: Java  
- Sinau | Software Developmet | Video 
 An android application that help mentee find various mentors nearby
 Semifinal at Gemastik’10 and Semifinal at Java Business Competition
 Tech stack: Java, Firebase  
