Projects
A detailed examination of key projects, outlining the challenges, processes, and outcomes.
Case Study 1: "Bersama Bill ID" AI Biller Identification System
Context
Artajasa, a leading payment infrastructure firm, required high-impact, innovative technology demonstrations for its annual "Members Meeting." This event is attended by key clients and stakeholders, including many executives from banks across Indonesia.
Challenge
With the event just about 2 months away at the time of my internship, a new series of compelling products was needed to serve for the flagship tech demo. The challenge was ours, the developers to continue developing and deploy the ideas from the remaining time left. One of the ideas lying around is "Bersama Bill ID", the idea is to pay bills using our face as the authenticator. This was assigned to me as the solo-developer for the optional purpose of POC for its annual Members Meeting should the task be done before the event.
Process
The project was executed in a series of intensive, week-long sprints, encompassing the full development lifecycle.
Foundational Learning & Architecture
The initial phase involved the rapid acquisition of unfamiliar technologies, including Go for backend microservices and Flutter for cross-platform mobile development. Simultaneously, the end-to-end system architecture was designed.
Core ML & Backend Development
The core machine learning model was developed with TensorFlow to intelligently identify customer faces as the authentication wall for billing payments. Concurrently, the supporting backend microservices and APIs were built using Go.
Mobile Frontend & Integration
The user-facing mobile application was developed using Flutter. This phase focused on creating an intuitive interface and ensuring seamless integration with the backend APIs.
Testing, Deployment & Polish
The final week was dedicated to system hardening. The application was containerized using Docker for stable deployment, followed by rigorous testing to ensure a faultless live demonstration.
Results
The final deliverable was "Bersama Bill ID," an AI-powered mobile application fully functional for the scopes of POC. The application was successfully demonstrated live on stage at the Members Meeting to hundreds of industry executives without technical failure.
I have successfully delivered a complete, full-stack, AI-driven POC product from concept to deployment in 1.5 months, working as the sole developer.
Case Study 2: IS 2024 Full-Stack Event Platform
Context
As the Head of Web Engineering for the IS 2024 university event, I was responsible for the digital infrastructure. The project was initiated during an accelerated summer semester, which created a high-pressure environment with a condensed timeline.
Challenge
The event required a centralized, custom-built web platform to manage all participant activities, from registration to scheduling. At the time I had no experience with fullstack programming, and no existing infrastructure was in place. The entire application had to be built from scratch against an extremely compressed deadline, with failure impacting all event participants.
Process
The process blended structured planning with intensive execution to meet the stringent deadline.
Architecture & Planning
The project began with system architecture design, including database schema, API endpoint definitions, and core user flow mapping. This initial planning was critical to prevent costly rework.
Intensive Solo Development
A two-week intensive development phase followed, where I single-handedly coded the majority of the backend logic and frontend interface to meet tightly timed initial critical milestones. This was done to prevent collaboration losses, as we all had no prior full-stack experience to know if we were heading into the right direction. During that time I rapidly reiterate experimenting back and forth to point no.1 and getting my hands dirty to get every basics right before further core feature integrations.
Integration & Testing
Core features were progressively integrated into a cohesive platform by our team. Continuous testing was also conducted to identify and resolve bugs, ensuring application stability.
Deployment & Handover
The platform was deployed to a live server. Essential documentation was prepared, and a handover session was conducted with event staff to empower them to manage the platform's operations.
Solution
The result was a robust, responsive, full-stack web application that served as the digital backbone for the IS 2024 event. The platform provided seamless participant's task submission and scoring by the committee, sophisticated attendance system, live information portals, and a full administrative control panel.
Results
The platform was launched successfully on schedule, with all core features fully functional.
The application reliably served hundreds of student participants throughout the event with no critical downtime and throttling.
This project solidified my full-stack capabilities and demonstrated the ability to deliver mission-critical software under extreme pressure.
Case Study 3: Applied ML in Geoscience: Low Resistivity Reservoirs
Context
A research initiative at the Faculty of Industrial Technology (FTTM) was investigating physics-based models for hydrocarbon reservoirs in low-permeability rocks. The project's analytical capabilities were constrained by the limitations of its existing data we had relevant to the topic.
Challenge
The research required a specialist to bridge the gap between theoretical physics and computational science. The core challenge was to translate complex physical equations into efficient Python code, develop improved data visualizations, and integrate predictive machine learning models to derive new insights from well logs regarding the probability of hydrocarbon reservoirs.
Process
Analysis & Enhancement
I began by conducting a thorough analysis of the existing python-based basic well log interpretation visualization models to identify areas for improvement.
Model Development
I applied various machine learning techniques to the reservoir data to build and test predictive models, aiming to enhance the model's forecasting capabilities.
Implementation
A key part of the process involved working directly with the professor to translate complex geophysical domain knowledge into functional, data-driven software solutions.
Results
The final deliverable included enhanced visualization tools and several predictive machine learning models applied to the reservoir data. The new models and visualizations provided the research team with enhanced tools for data analysis and prediction.
Case Study 4: "Kuliah Kit" Efficient College Learning Platform
Context
College students frequently face the challenge of managing dense lecture materials and tracking their understanding across multiple courses. Traditional study methods often lack a structured way to prioritize topics based on mastery. Based on extensive user interviews, students expressed a clear desire: "I want to understand materials quickly without the hassle." They need a way to track progress, know what to study next, and keep their files organized.
Challenge
The primary challenge was to design a digital solution that transforms static lecture slides into an interactive study system. The platform needed to leverage Large Language Models (LLMs) to intelligently break down materials, assess their confidence levels for each topic, and visually track their progress to optimize study sessions. A critical requirement was ensuring the accuracy and reliability of the AI-generated content, as educational hallucinations could be detrimental to the learning process. Key requirements included speed (smooth upload/processing), clarity (instant summaries & goals), assessment (meaningful quizzes & metrics), and retention (progress tracking & weak concept reminders).
Process: The User Flow
The solution was architected around a streamlined user flow designed for efficiency and retention, focusing on the core loop of Upload → Learn → Assess. Aiming to answer our value proposition: "Learn faster, understand better."
1. Onboarding & Personalization
A frictionless entry asking key details to tailor future recommendations.
2. The Core Activity Hub
The Home page acts as the central command center. Focused on a clean, distraction-free interface that prioritizes content. The design centered around the concept of "material analysis" where users can tag topics as understood or needing review.
3. Seamless Upload Experience
A critical "Upload Flow" designed to be straightforward and smooth. Users can drag & drop PDF, PPT, DOC, or any other lecture files. A processing screen keeps the user informed ("KuliahKit is summarizing, mapping goals...") while the LLM pipeline analyzes and structures the material. We implemented rigorous prompt engineering to maximize the fidelity of the extraction process.
4. The Learning Engine (Material Page)
The heart of the application, powered by generative AI. This page delivers:
- Summary: AI-synthesized 5-point quick summary and a 1-paragraph overview, optimized for factual accuracy.
- Learning Goals: 3-6 actionable bullet points to orient the user's focus.
- Quiz: LLM-generated 5-question assessment with immediate feedback and explanations, cross-referenced with the source material.
- Performance Metrics: Visualizing "Strengths" (concepts mastered) and "Weaknesses" (areas to review).
- Practice Suggestions: Concrete ways to apply the knowledge.
Results & Validation
User validation confirmed the effectiveness of this flow. Users specifically praised the Summary and Quiz features for saving time.
Actionable Insights: The "Strengths & Weaknesses" analysis provided the "what next?" guidance students craved.
Retention: The structured navigation (Recent, Bookmarks) and progress tracking successfully addressed the need for an organized, gamified learning environment.
Case Study 5: "Streak" Habit Tracking & Personal Growth
Context
Consistency is the engine of personal growth, yet maintaining new habits is notoriously difficult. Many people understand the theory but lack a practical tool to visualize their consistency and maintain momentum.
Challenge
The challenge was to build a habit tracking application that goes beyond simple checkboxes. It needed to educate users on the principles of behavioral change while providing a compelling visual representation of their "streaks" to gamify the process of self-improvement.
Process
Conceptualization
Researched habit formation psychology to design features that reinforce positive behavior. The core concept was built around the "streak" mechanic to leverage loss aversion as a motivator.
Implementation
Developed a responsive web application using Next.js. Key features included a dashboard for daily tracking, visual graphs of progress over time, and educational resources explaining the "why" behind habit formation.
Results
Streak serves as a digital companion for personal development. It successfully combines utility with educational content, helping users not just track what they do, but understand how to change their behavior effectively.