MUHAMADGALIH
Engineer by day, illustrator by night. I make the web a little less boring.
Projects shipped
Years across code & canvas
Clients who came back
Cups of coffee, uncounted
Full-Stack Engineer, UI/UX Designer, Illustrator, and Data Scientist & Analyst based in Indonesia.
Skills & Tools
What's open in my editor and Figma tabs on any given day, from first sketch to shipped code.
Design & Creative
Adobe Photoshop
Adobe Illustrator
Canva
CapCutDevelopment
SQL
Power BI
Pandas
NumPy
Microsoft Office
Illustration & Graphic DesignMini Tournament NOTARY CUP JATENG
Developed the complete visual identity and design assets for the Notary Cup JATENG - Mini Padel Tournament, creating a sophisticated monochrome brand that blends the professionalism of the legal field with the dynamic spirit of padel. Produced over 20 design assets, including social media content, tournament medals, venue banners, and custom ribbons, ensuring a cohesive visual experience that strengthened the tournament's branding and participant engagement.
Illustration & Graphic DesignMarii Club x HPC Golden Hour Padel Session by Momentum
Designed a comprehensive visual identity and 20+ promotional assets for the Marii Club x HPC Golden Hour Padel Session organized by Momentum. Deliverables included event branding, backdrop banners, interactive Instagram feeds/stories, match schedules, and player cards. By applying a modern, sleek, dark-themed aesthetic across both venue setups and social media graphics, I delivered a unified high-end brand experience that maximized event visibility and participant excitement.
Illustration & Graphic DesignMini Tournament Kangkung Bakar by Momentum
Crafted the visual identity and created 20+ design assets including the official tournament logo, event venue banners, and interactive social media feeds/stories for a 40 player (20 teams) Padel tournament. By developing a clean and energetic visual framework for group schedules and promotional posts, I ensured strong brand consistency across both digital channels and the venue, directly elevating participant engagement and tournament visibility.
UI/UXFoldIn.space - UI/UX Redesign & Design System
Led the end-to-end redesign of FoldIn, a social network for local events, building a token-driven Figma design system (~92 tokens, 6 collections, 5 breakpoints) that unified a fragmented UI from 8 fonts into one cohesive identity. Designed a 25+ component library, a custom flat icon system, and 50+ responsive screens spanning onboarding, events, checkout, wallet, and settings/KYC. Conceived and designed the flagship Ghost Match feature, turning verified event attendance into a connection-discovery experience.
UI/UXend-to-end UI/UX enhancement of the Selasar Sunaryo Art Space website
Led a UI/UX enhancement of the Selasar Sunaryo Art Space website, focusing on making its exhibitions, events, and programs easier to discover and navigate. The work covered restructuring the information architecture, refining the visual hierarchy and typography to match the gallery's artistic character, and improving key user flows such as browsing current and past exhibitions, viewing event schedules, and finding visit information. The redesign was prototyped in Figma with a responsive, mobile-first approach, resulting in a cleaner interface, more consistent components, and a more immersive experience that better represents the space online.
SoftwareBakery Shop - Bakery ERP System
Bakery ERP System is a full-featured Enterprise Resource Planning application built to manage the end-to-end operations of a bakery business. Developed with Laravel 12, Filament 4, Tailwind CSS, and Alpine.js, it unifies procurement, production, inventory, sales, and accounting into a single admin platform. The system covers the complete supply chain, from purchase orders, vendor management, and goods receipts, through recipe-driven production using Bill of Materials (BOM) and manufacturing orders, to real-time inventory tracking via stock movements and adjustments. On the sales side, it includes an integrated Point of Sale (POS), sales order processing, and a customer loyalty engine featuring tiered memberships, a points ledger, and configurable discount rules. A built-in double-entry accounting module with a Chart of Accounts, journal entries, and vendor payments ensures every transaction is financially reconciled, while role-based access control (Filament Shield), PDF invoicing, and Excel import/export round out a production-ready, business-grade solution.
Graphic Designer at PT Bukhori Indonesia
Managed the design needs of three company brands (Nusa MUN, Leaders, and Youthspace Indonesia) while maintaining the consistency of each brand's visual identity. Designed promotional materials for 5+ competitions and events, averaging around 10 assets per event, covering Instagram feeds, posters, and competition guidelines. Prepared merchandise assets for competitions with up to 1,000 participants, including t-shirts, medals, certificates of achievement, certificates, and stickers.
ANALISIS SENTIMEN FENOMENA #KABURAJADULU PADA PLATFORM X MENGGUNAKAN LONG SHORT-TERM MEMORY (LSTM)
This study classifies public sentiment toward the #KaburAjaDulu hashtag phenomenon on Platform X using a Deep Learning approach with an LSTM architecture. The dataset consists of 19,572 clean tweets (from 19,848 raw) scraped with Python and Selenium over the January to May 2025 period, then auto-labeled with a lexicon-based method and vectorized through token sequencing and padding. Fifteen experiment combinations (three split ratios times five hyperparameter scenarios) were tested. The best configuration, namely embedding 256, 256 LSTM units, 15 epochs, and a 70:30 split, achieved 98.35% validation accuracy with a stable learning curve and no overfitting. The sentiment distribution was dominated by Neutral (65.65%), followed by Negative (28.13%), which was 4.5 times more frequent than Positive (6.22%). Supported by word cloud analysis, the findings confirm that the phenomenon is a form of critical reflection and escapism against structural pressures rather than a mere humor trend.
UI/UXDaylight Asia - UI UX Design
Sistem Pembelajaran Siswa Sekolah Dasar
Digital Learning Media for Elementary School is a web-based educational platform built with Laravel and Livewire, designed to support elementary school teaching and learning. The platform provides role-based dashboards for administrators, teachers, and students, enabling teachers to manage and deliver interactive learning materials while students access lessons through an intuitive, responsive interface. With a clean and friendly design tailored for young learners, the application aims to make classroom content more engaging and accessible both in and outside the classroom.
Implementasi Algoritma Genetika untuk Penentuan Komposisi Portofolio Saham Optimal dengan Pendekatan Multi-Objektif
This study applies a Genetic Algorithm (Real-Coded GA) to determine the optimal stock portfolio composition, aiming to balance risk and return. The fitness function is designed as a linear combination of four financial metrics (Sharpe, Calmar, Sortino, and Information Ratio), using data from 10 Indonesian Stock Exchange (IDX) bank stocks pulled from Yahoo Finance (yfinance) since January 2020. The GA uses tournament selection, extended intermediate crossover, reciprocal exchange mutation, and 10% elitism. As a result, the optimal portfolio reached a fitness score of 1.8709 compared to 1.3886 for the equal-weight portfolio, a 34.73% improvement, with a Sharpe Ratio of 1.46, Calmar of 2.09, and an annual return of 32%. Configuration experiments showed that population size has a greater impact than the number of generations, with the combination of 200 population and 500 generations delivering the most stable convergence.
Pemodelan Pengguna Spotify Menggunakan Machine Learning: Reduksi Dimensi, Klasterisasi, dan Analisis Aturan Asosiasi
Analisis ini berhasil memetakan karakteristik pengguna, mulai dari kelompok Light Listener hingga Heavy Listener, serta menemukan pola unik seperti keterkaitan kuat antara genre Folk, Electronic, dan Metal pada kelompok tertentu. Informasi ini dapat digunakan Spotify untuk mengoptimalkan sistem rekomendasi dan strategi pemasaran yang lebih personal.
Gold Price Prediction with KNN and Linear Regression
Projek ini merupakan studi kasus machine learning oleh Muhamad Galih yang bertujuan untuk memprediksi pergerakan harga emas (Gold Futures) menggunakan dataset historis dari Yahoo Finance periode 2015 hingga April 2025. Penelitian ini menerapkan dua pendekatan utama, yaitu klasifikasi menggunakan algoritma K-Nearest Neighbors (KNN) untuk menentukan arah harga (naik/turun) dengan akurasi terbaik sebesar 89,12%, serta regresi linear untuk memprediksi nilai numerik harga penutupan dengan performa $R^2$ mencapai 0,9958. Melalui proses feature engineering yang melibatkan indikator teknikal seperti Moving Averages (MA), Exponential Moving Averages (EMA), momentum, dan volatilitas, model ini berhasil memberikan prediksi harga emas untuk tanggal 24 April 2025 sebesar $3405.36 dengan keyakinan arah harga akan naik sebesar 100%.
Pareeduhub - Online Course Platform (LMS)
- Engineered a full-stack online course platform (Laravel 12, React, Inertia.js, TypeScript, MySQL) with course management, enrollment, and progress tracking. - Integrated Midtrans & Flip payment gateways and built secure video streaming with watermarking and anti-piracy protection.
Optimalisasi Pemenuhan Gizi dengan Algoritma Genetika untuk RDA
Menjelaskan implementasi Algoritma Genetika untuk menemukan solusi optimal dalam pemenuhan Angka Kecukupan Gizi (AKG) berdasarkan dataset makanan tertentu. Model ini mengevaluasi berbagai kombinasi makanan untuk memenuhi kriteria nutrisi seperti kalori, karbohidrat, protein, lemak, vitamin C, dan kalsium dengan batasan total berat dan biaya maksimal. Melalui proses evolusi yang mencakup seleksi (seperti roulette wheel, tournament, atau rank), one-point crossover, dan insert mutation, algoritma ini mampu mencapai kondisi konvergensi untuk menghasilkan daftar bahan makanan dengan nilai fitness terbaik. Berdasarkan pengujian dengan konfigurasi berbeda, penggunaan metode Tournament Selection menunjukkan performa paling unggul dalam mencapai nilai fitness tertinggi dibandingkan metode lainnya.
PENGEMBANGAN SISTEM OTOMATISASI WEB SCRAPING DAN RINGKASAN BERITA MENGGUNAKAN FLASK, NLTK, DAN PENDEKATAN METODE TF-IDF
This study builds a web-based system to automate the collection and summarization of news from various Indonesian online portals. The system uses BeautifulSoup and Requests for scraping, NLTK and Sastrawi for text preprocessing (tokenization, stopword removal, stemming), and the TF-IDF method to weight words and rank sentences into an extractive summary. The interface is built with Flask, allowing users to set the summary percentage in real time. Tested on 10 articles from different portals with a 30% summary length, the system achieved an average ROUGE F1 score of 57%, with high precision (close to 1) despite lower recall due to the short summary length. The results show that the system preserves the essence of the main information even when articles are significantly condensed.
SiTubel (Sistem Tugas Belajar) — Employee Further Study Management System
Developed a full-stack web application using Laravel and Filament to digitize the further-study (tugas belajar) application and approval process for civil servants at a local government office. Designed a role-based multi-panel architecture (Admin, OPD, and Employee panels) with custom authentication using Employee ID (NIP) instead of email, and implemented granular role and permission management via Spatie Laravel-Permission. Built core modules for employee (pegawai) data management, organizational unit (unit kerja) administration, and a multi-stage application workflow that tracks each submission through OPD review, BKPSDM verification, selection, and graduation stages with real-time status updates. Integrated document upload and media management (Spatie Media Library) for handling required administrative files, and designed the database schema and Eloquent relationships to support scalable data across employees, study programs, and institutions.
Optimize Random Forest Classification Performance on Lung Cancer Data Imbalance with SMOTE Method
This study addresses the class imbalance problem in a lung cancer dataset (238 negative vs. 38 positive after duplicate cleaning) using the Synthetic Minority Over-sampling Technique (SMOTE). SMOTE was applied only to the training data to keep the evaluation objective, and classification was then performed using Random Forest. As a result, all metrics improved: accuracy 87.5%→91.1%, recall 88%→91%, precision 87%→92%, and F1-score 86%→90%. This demonstrates that SMOTE effectively improves model performance in medical diagnosis cases with a small minority class.
Implementasi Fuzzy SAW dalam Penentuan Dosen Pembimbing Kerja Praktik
Dalam simulasi kasus mahasiswa dengan fokus bidang Rekayasa Perangkat Lunak (RPL), metode ini berhasil memberikan perangkingan alternatif, di mana Alternatif A3 memperoleh nilai preferensi tertinggi (2.49) dan terpilih sebagai alternatif terbaik.
Kind words from collaborators and clients I've had the pleasure of working with.
“Desain Baju Ilustrasi dari beliau cukup bagus dan sangat memuaskan.”
Ifan Maul
Fullstack Engineering @ Freelance
“Sangat keren sekali, terimakasih✨✨”
Adinda Oktaviani
Have a project in mind, want to collaborate, or just want to say hi? My inbox is always open.
© 2026 Muhamad Galih · MIZARIE · All rights reserved
MG / MIZARIE