★ STRICTLY CONFIDENTIAL ★
RESEARCH & ENGINEERING LAB DOSSIER
Software Engineer · Mobile & Web · Applied AI
⚡ 90.1% Saved
Label Cost
First Author · IEEE '26
📱 Production
App Ecosystem
Arthawira B2B Fleet
🎓 GPA 3.95
Summa Cum Laude
STMIK & Vocational Lead
"Candidate Dossier Verified & Ready for Inspection 🐾"
Select a direct channel or enter portfolio below:
Open to software engineering opportunities
I build practical mobile and web applications that solve real problems.
I build mobile and web applications with Flutter and Laravel, with a research background in NLP and applied machine learning.
Three Professional Proof Pillars
· how engineering is differentiatedEvidence-first profile
01 — BUILD (CORE)
PrimaryMobile & Web Applications
Flutter · Laravel · REST API · State Management
02 — RESEARCH
NLP & Active Learning
IndoBERT · Active Learning · FAHMA / ICERA
03 — COMMUNICATE
Technical Teaching
SMK Koperasi · 140+ Students · Curriculum Design
01 what's up
"Building practical mobile and web systems, backed by applied NLP research."
I build practical, production-ready software designed for longevity and real user value. My core engineering toolkit centers on Flutter for responsive mobile applications and Laravel for structured backend architectures, supported by offline-first SQLite persistence and clean Git workflows.
Beyond shipping software, I conduct applied research in Natural Language Processing and active learning (IndoBERT) to optimize annotation efficiency. Teaching Coding and Artificial Intelligence to 140+ vocational students at SMK Koperasi also honed my ability to translate complex architectures into clear, actionable engineering principles.
02 project dossiers
click any folder to inspect complete technical dossier
A curated collection of mobile and web applications I have built—focusing on offline-first architectures, API integrations, and maintainable systems.
Engineered to resolve financial privacy concerns and eliminate transaction query latency without relying on continuous cloud connectivity.
A cultural heritage preservation system utilizing computer vision to digitize traditional Javanese script in real-time.
💡 Tip: Click any folder flap above or the numbered tabs #01 through #06 to open the full technical specification dossier.
03 scrapbook pinboard desk
hover paper to lift & inspect
A tactile pinboard inventory of production-tested mobile frameworks, web architectures, machine learning toolkits, and verified licenses.
Building resilient, offline-first cross-platform applications with native performance and responsive UI.
Architecting robust web platforms with normalized relational databases, clean Blade templates, and RESTful APIs.
Empirical machine learning and Indonesian transformer models.
★ Active Learning & Annotation Efficiency
Clean version control, API testing, and UI design tokens.
Strict PR Reviews & Systematic Debugging
Verified industry licenses, technical trainings, and IEEE accreditations.
04 experience & leadership
practical track record
A progressive record of production software engineering, dedicated vocational coding instruction, and academic community leadership.
Client & Production Deployments
Designing and shipping production mobile apps (Flutter) and web platforms (Laravel), with end-to-end product design, UX mapping, and QA validation for client and institutional systems like SIMPEG El-Rahma, Arthawira, and Seulanga Kost.
SMK Koperasi Yogyakarta
Grade 10: Coding & AI (KKA) · Grade 12: Figma UI/UX Fundamentals
Permikomnas Yogyakarta
Collaborated in rapid prototyping for Garuda Hackathon 4.0, contributing to front-end UI implementation, basic interaction design, and design thinking problem solving.
Field Logbook #04
Engineering Dividends
Mentorship & Delivery Synergies
Why this benefits software teams:
"Teaching code proved that true technical mastery is demonstrated by the ability to explain complex abstractions simply."
05 empirical lab dossier
scientific rigor
A specialized research focus on transformer language representations and annotation-efficient active learning, validated through peer-reviewed academic publications and reproducible empirical benchmarks.
Lab Dossier #05 · IndoBERT & Multi-Oracle Specialization
Applied machine learning & NLP research focused on data annotation cost reduction through an Adaptive Multi-Oracle Active Learning architecture. Combines IndoBERT contextual representations with entropy-based uncertainty routing across 3 label oracles: Human Annotator (weight 1.0), Pseudo-labels IndoBERT (weight 0.3), and Fine-tuned IndoBERT (weight 0.7).
90.1%
Label Cost Saved
149 queries to hit 90% benchmark performance
~8%
Human Burden
Human attention spent only on ambiguous samples
0.6277
Macro-F1 Score
Evaluated across 8,966 YouTube MBG comments
2 Papers
Peer-Reviewed
IEEE ICERA 2026 & Jurnal FAHMA (Sinta 4)
Peer-Reviewed Academic Credentials
·primary author & IEEE accepted
★ IEEE International Conference on Electronics, Robotics and Automation · IEEE Xplore
Minarwati and A. P. Hardiadi (2026).
"Adaptive Multi-Oracle Active Learning using IndoBERT Representations for Efficient Indonesian Sentiment Classification"
IEEE (ICERA 2026).
DOI: 10.1109/ICERA72709.2026.11666714
Core Contribution: Designed an adaptive 3-oracle active learning pipeline with entropy-based query routing, cutting human annotation costs by 90.1% using 768-d embeddings from Hugging Face model Aardiiiiy/indobertweet-base.
★ Jurnal Informatika Komputer, Bisnis dan Manajemen · Vol. 24 No. 2 (2026)
Alvian Putra Hardiadi and Minarwati (2026).
"Analisis Active Learning SVM berbasis Margin Sampling pada Sentimen YouTube MBG"
Jurnal FAHMA (Sinta 4).
DOI: 10.61805/fahma.v24i2.203
Core Contribution: Proved that the Human-in-the-Loop *skip* mechanism acts as an implicit quality filter preventing noisy annotations from entering the training set, achieving a +0.0235 Macro-F1 gain with only 50 annotated samples.
Technical Methodology & Empirical Evidence
·architecture & verified plots
Iterative query selection loop routing YouTube comments to the optimal oracle based on model prediction uncertainty (entropy):
Unlabeled Pool: 8,966 Comments
Public YouTube comments on Indonesia's Nutritious Meal Program
IndoBERTweet Feature Extractor
Embedding 768-d L2-normalized [CLS] tokens
Entropy-Based Uncertainty Routing
Adaptive Weighted LinearSVC
90.1% manual annotation cost reduction achieved
Experimental Plots
Figure 1 (IEEE ICERA 2026): Macro-F1 Progression across 5 active learning strategies over 1,500 queries. Multi-Oracle configuration reliably converges to top benchmark.
let's chat
SET A CALL, A PROJECT, OR JUST WANT TO SAY HI? SEND IT OVER! I READ EVERY MESSAGE.
* Open to software engineering roles, hybrid or remote
