MUHAMAD KEMAL FAZASELECTED WORK / 2026

Building systems.
Understanding people.
Connecting both.

I’m Kemal, an Informatics student at Diponegoro University. I build digital products and keep asking how the systems behind them work, and who ends up using them.

Explore my work

A FEW THINGS I’VE BUILT

Selected work(04)

Different problems.
The same curiosity.

01 / PRODUCT ENGINEERING

YoDips.

Three academic systems.
One student experience.

YEAR / CONTEXT: 2026 · Self-initiated · Individual
MY ROLE: Sole Developer & Product Owner

Less fragmentation. More focus.

Assignments, grades, schedules, and attendance all sit in separate UNDIP systems. I built YoDips to bring SSO, Kulon, and SIAP into one place, across web, Android, and a browser extension. A few close friends use it too.

Behind the build

Convenience should not cost you your password.

Users sign in with the university directly. YoDips reuses the verified session instead of collecting or storing their password.

YoDips browser extension setup with the backend URL redacted

What I learned: architecture and UX are connected. A secure system that feels confusing still creates friction.

02 / ACCESSIBILITY

LevelUp.

Different minds.
Different ways to learn.

YEAR / CONTEXT: 2026 · Competition · Team of 3
MY ROLE: Project Manager, UI Baseline Designer
& Contributing Developer
RECOGNITION: ↗ CITECH 2026 Finalist · Universitas Jember

Learning that adapts to the learner.

LevelUp changes its interface to fit how someone reads and focuses, with profiles for ADHD, Dyslexia, and general learning. I coordinated the team, set the UI baseline, and built a few of the features myself.

Explore the approach

Reading controls include Lexend and OpenDyslexic, adjustable font size and spacing, text-to-speech, and focus tools. Quizzes also adjust their difficulty as you go.

CITECH 2026 finalist certificate — Finalist evidence · CITECH 2026

What I learned: accessibility is a design decision from the start, not a settings page you add at the end.

03 / HUMAN–COMPUTER INTERACTION

LastBite.

Good food.
A second chance.

YEAR / CONTEXT: 2026 · Class + Competition · Team of 3
MY ROLE: Programmer during the team prototype;
built it out on my own afterward
VALIDATION: 11 users · System Usability Scale

From class prototype to something I kept building.

LastBite connects businesses that have safe, unsold food with people looking for a cheap meal. Buyers browse discounted listings and pick up at a scheduled time. After the group prototype, I kept going and built it out further on Android.

What testing taught me

We tested the prototype with 11 users using the System Usability Scale (SUS).

A flow that makes sense to me does not always make sense to the user. Building past the assignment showed me how much a prototype leaves open.

04 / MACHINE LEARNING

Beyond the pixels.

Deepfake classification.
Evidence over intuition.

DAC Find-IT Kaggle notebook and submitted private score of 0.73969

PROJECT: DAC Find-IT 2026
YEAR / CONTEXT: 2026 · Competition · Team of 3
MY ROLE: Collaborative ML Experimentation

Getting a good score is an experimentation problem.

Each of us trained and compared approaches. Our submitted notebook used ConvNeXt Tiny for six-class classification, stratified five-fold validation, and a five-model ensemble.

0.8177Best fold validation Macro F1

0.73969Kaggle private score

Inside the experiment

We used class-balanced weights, cross-entropy with 0.1 label smoothing, and CosineAnnealingLR. We saved the best model per fold and ran three-way test-time augmentation, giving 15 probability outputs per image.

What I learned: validation, class imbalance, augmentation, and inference decisions affect reliability as much as architecture does.

THE PERSON BEHIND THE PROJECTS

From “does it work?”
to “who is it for?”

My projects slowly moved from pure implementation toward product thinking, accessibility, user testing, and experimentation. The tools changed. The habit of learning did not.

I’m studying Informatics at Diponegoro University. I treat systems, accessibility, and machine learning as different tools for the same job: understanding a problem and solving it.

Understand

Frame the real user or system problem.

Build

Turn an idea into something concrete.

Test

Look for evidence, not assumptions.

Learn

Take what worked into the next version.

THE NEXT CHAPTER / APPLE DEVELOPER ACADEMY

I know how to build. Now I want to get better at deciding what is worth building, by learning from users and working with people who think differently from me.

LET’S CONNECT

Good things start
with a conversation.

kemalfaza26@gmail.com

Building systems.Understanding people.Connecting both.