Software engineer · Bangkok, Thailand

I build practical software and language technology.

I’m Aung Thu Hein, a software engineer working across full-stack product systems and open-source Myanmar NLP.

Explore selected work ↘Résumé ↗GitHub ↗
Black-and-white portrait of Aung Thu Hein
01 Engineer, student, and builder of tools for Myanmar language computing.

Selected work · 2024–2026

A few projects,
shown with intent.

Product systems, language tooling, and applied ML—selected for the decisions and outcomes behind them.

01 / Professional systems

New Generation Myanmar

Turning academic operations into dependable digital workflows.

I contribute across training-management and student-information platforms, translating domain rules into maintainable interfaces, services, permissions, and structured data flows.

  • ContextReal users, operational constraints, and role-aware access
  • ContributionFull-stack features from workflow definition to delivery
  • FocusConsistency, maintainability, and clear business rules
Visit organization ↗
New Generation MyanmarOfficial homepage ↗
02 / Open source

BurmeseNLP

A dependency-light toolkit for Myanmar-language processing.

I designed a staged, deterministic Python pipeline for normalization, Zawgyi conversion, segmentation, POS tagging, phrase chunking, NER, clause parsing, and corpus export—without mandatory model downloads.

8pipeline capabilities
0required model downloads
PyPIreusable package

Documentation ↗Source code ↗

pipeline.py
from burmesenlp import Pipeline

nlp = Pipeline([
  "normalize",
  "segment",
  "pos",
  "ner"
])

result = nlp("မြန်မာစာ")
03 / Applied ML

Laya Burmese SIB-200

Making Burmese topic predictions more accurate—and more honest.

I fine-tuned a 0.3B multilingual model for seven-class Burmese topic classification, then applied temperature scaling so confidence scores better reflected real performance.

71.6%test accuracy
67.7%macro-F1
0.051calibration error

Try the model ↗Model card ↗

Test accuracy

Official 204-example SIB-200 test split
Burmese RAG logo

Local AI · Retrieval

Burmese RAG

A Zawgyi-safe, syllable-aware retrieval system for chatting with Burmese documents. It combines multilingual embeddings with syllable n-gram BM25, supports local or hosted LLMs, and returns cited answers.

View repository ↗
Chart comparing the Burmese token cost of popular AI model tokenizers

Research benchmark · LLM efficiency

Myanmar Tokenizer Bench

A reproducible FLORES-200 benchmark showing that popular AI tokenizers use 1.6× to 11.7× more tokens for Burmese than English—affecting cost, speed, and usable context.

Explore benchmark ↗
က

Browser inference · Co-developed

Learn Myanmar Script

An educational handwriting experience running ONNX inference directly in the browser, with drawing, feedback, and quizzes—no server-side model API.

View project ↗
MarkdownEditor logo

Desktop engineering

MarkdownEditor

A packaged Java editor that keeps filesystem state, editing history, and a JavaFX WebView preview synchronized, with GFM rendering and PDF export.

View project ↗
More experiments and earlier work +
KryptoJavaFX · EncryptionGEDSimAI assessmentService ManagementReact · Spring BootEchoPostLaravel · ReactTrash TradeFlask · CVDevAskJava EE · MySQL
02 / About

Engineering with a reason behind it.

I started by building web and desktop applications, then became interested in a problem that affected the tools I wanted to create: Myanmar language support was still limited.

Today, I work on education platforms professionally while developing NLP libraries, models, and browser-based learning tools independently. I enjoy projects where product engineering and applied machine learning meet.

I care about systems that remain understandable after the demo: explicit domain rules, testable components, useful documentation, and model results presented with the right context.

Working mainly with Python, Java, TypeScript, React, Laravel, and modern NLP tooling.

Experience & credentials

Grounded in delivery.
Still learning in public.

Software Engineer

New Generation Myanmar

Full-stack education platforms and operational workflows

Bachelor of Computer Science

Rangsit University

Software engineering and applied AI

ITPEC Fundamental IT Engineer

FE Certification

Algorithms, systems, networks, databases, and security

ITPEC Information Technology Passport

IP Certification

Broad computing and technology foundations

Have a useful problem?

Let’s build something
that matters.

I’m open to software engineering, applied AI, open-source collaboration, and thoughtful conversations about Myanmar language technology.

GitHubHugging FaceLinkedInThreadsMedium