Back to Home
PROJECTS ARCHIVE

A long-form archive rendered from Markdown

Automation
WhatsApp Expense Tracker v2
From bloated Node + Chromium to a lean Go webhook architecture
12/23/2025
0 views
GoWA handles WhatsApp connectivity; a Go listener/parser processes events and syncs to Google Sheets via webhooks.

GoWA handles WhatsApp connectivity; a Go listener/parser processes events and syncs to Google Sheets via webhooks.

Summary

This is a v2 rewrite of my WhatsApp expense automation focused on reliability and operations. The old build worked, but it was too heavy: a single Node.js service running WhatsApp Web automation through Puppeteer + Headless Chrome.

v2 introduces a decoupled, webhook-driven architecture and moves the core logic to Go, cutting resource usage while making the system easier to run long-term on a home lab.

techstack
Go · GoWA · Webhooks
Google Sheets API · GCP Cloud APIs (OCR/LLM)
Docker · Grafana (cAdvisor)
Same user value, radically simpler runtime footprint.
Infrastructure
Utilizing Tailscale for Self-host/Homelab Setup
Zero setup VPN and Private (well not fully)
10/10/2025
0 views
Tailscale turns your scattered devices into a secure, private mesh network—perfect for linking homelabs, laptops, and mobile clients seamlessly across locations.

Tailscale turns your scattered devices into a secure, private mesh network—perfect for linking homelabs, laptops, and mobile clients seamlessly across locations.

Summary

A personal initiative to build a secure, globally accessible homelab infrastructure interconnecting multiple physical locations (Jakarta and Yogyakarta) and various devices using Tailscale. The setup enables seamless, encrypted access to private and public web services without exposing them directly to the internet.

Manage and connect private connection with ease.
Automation
AI WhatsApp Expense Tracker
Track your expenses by just chatting
9/26/2025
0 views
Send a chat with a receipt image — AI reads it, logs it, and updates your spreadsheet.

Send a chat with a receipt image — AI reads it, logs it, and updates your spreadsheet.

Summary

Born from frustration with messy financial habit tracking, this project reimagines expense management through everyday chat. Instead of manually logging every purchase, users simply send a quick message or upload a transaction photo to their private WhatsApp chat. The system automatically extracts key details, categorizes the expense, and syncs it directly to Google Sheets for analysis.

techstack
Node.js · wweb.js · Gemini API
Docker · Grafana · Google Sheets API
Home Lab Deployment · Privacy-first LLMs
Built in a home lab as a weekend experiment, currently scaling into production ready app.
Computer Vision
Web OCR System
Automating Visual Data Entry and Recognition with AI
10/10/2024
0 views
A custom-built OCR pipeline leveraging deep learning for real-world numeric and text recognition tasks on utility meters and similar image datasets.

A custom-built OCR pipeline leveraging deep learning for real-world numeric and text recognition tasks on utility meters and similar image datasets.

Summary

A freelance project aimed at automating visual data entry from numeric meters (e.g., electricity readings) using machine learning-based Optical Character Recognition (OCR). This system tackles the repetitive and error-prone process of manually transcribing blurred or low-quality meter images by introducing an AI pipeline that reads, processes, and interprets images directly from a web interface.

techstack
Roboflow · YOLOv8 · Python (Streamlit)
AI-assisted visual data automation for real-world applications.
Machine Learning
Web-Based Javanese Script Classifier
Preserving Cultural Heritage Through AI-Powered Handwriting Recognition
8/30/2024
0 views
A web-based deep learning system for classifying handwritten Javanese script characters, enabling users to learn and explore traditional scripts interactively.

A web-based deep learning system for classifying handwritten Javanese script characters, enabling users to learn and explore traditional scripts interactively.

Summary

A research-driven project that applies machine learning to preserve the Javanese script, one of Indonesia’s historical writing systems. This system allows users to draw Javanese characters on a web canvas and receive real-time classification predictions, promoting cultural literacy through interactive technology.

This project served as the implementation of a Bachelor’s thesis, focusing on leveraging transfer learning for handwriting recognition in underrepresented scripts.

Transfer learning meets cultural heritage preservation.
Machine Learning
Techwaste
A Smart Solution for Managing Electronic Waste with AI
9/10/2023
0 views
An AI-powered platform that classifies and manages electronic waste, promoting recycling and sustainable disposal in Indonesia.

An AI-powered platform that classifies and manages electronic waste, promoting recycling and sustainable disposal in Indonesia.

Summary

Techwaste is a machine learning-based application that helps users identify, categorize, and manage electronic waste (e-waste). Developed during the Bangkit Academy 2023 batch, this project tackles one of Indonesia’s fastest-growing environmental challenges: electronic waste accumulation.

By combining AI-powered classification with cloud architecture, Techwaste streamlines recycling awareness and promotes responsible disposal through digital innovation.

Redefining sustainability with machine learning and cloud technology.
Web Development
Web-Based Information System
Empowering NGOs with Real-Time Volunteer and Disaster Management
11/15/2022
0 views
A centralized CMS platform designed for NGO stakeholders to manage volunteers and disseminate real-time disaster information efficiently.

A centralized CMS platform designed for NGO stakeholders to manage volunteers and disseminate real-time disaster information efficiently.

Summary

A full-stack web-based information system designed for an NGO (YAKKUM Emergency Unit) focusing on natural disaster response. The project aimed to streamline volunteer management, activity reporting, and information distribution, ensuring stakeholders and the public have access to real-time, verified updates during crisis events.

This project was developed as part of an internship program, completed by a team of five developers, and served as a fifth-semester academic project.