LWIN MOE AUNG
0
Lat 1.3521° N Lon 103.8198° E Bangkok / Singapore · 2026

Data Scientist · ML Engineer · AI Specialist

Lwin Moe Aung

Building systems that reason. Models that move. Stories that data alone could not tell.

Portfolio — 02 / 26
Scroll

01 / Manifesto

I   t u r n

data into decisions,

models into momentum,

insights into action.

$1.6M at risk · AUC 0.86 · 607K+ records · Silhouette 0.891 · $12.6M analysed · 5K+ respondents · RMSE −50% ·

02 / Selected Work

Six scenes,
one practice.

Freelance 2026

Customer Churn
Prediction.

XGBoost · SHAP · Streamlit · 7,043 customers

Built a full 10-phase churn prediction pipeline on 7,043 IBM Telco customers — ETL, data wrangling, feature engineering, training, and evaluation. XGBoost achieved AUC 0.86 on holdout with zero label leakage, confirmed by a SHAP interpretability audit, deployed as a 5-tab live Streamlit dashboard. Quantified $1.6M in annualised revenue at risk from the month-to-month base and flagged the M2M × Fiber Optic segment (2,128 customers, 54.6% churn, ~$11M ARR) as the highest-priority retention target. Delivered three ROI-quantified interventions totalling $1.5M+ in defended revenue. Surfaced "Dependents" as SHAP feature #2 — a 3.9× churn spread invisible to univariate EDA.

0.86AUC holdout
$1.6Mrevenue at risk
$1.5M+defended
XGBoostSHAPStreamlitPythonDuckDB
View Project →
01
Freelance 2025

Personal AI
Assistant.

NLP & Agentic System · OpenClaw + Ollama

Built and deployed a locally hosted NLP-powered agentic AI assistant using OpenClaw and Qwen3.5 397B via Ollama — zero API cost, full local deployment — autonomously handling text processing, task management, research, and document drafting. Engineered a multi-step reasoning pipeline with tool-use, contextual memory, and text preprocessing modules; deployed as a local API integration between the model and application layer. Currently extending to outbound messaging automation and ETL-style data pipelines.

Qwen3.5397B parameters
$0API cost
Localfull control
OpenClawOllamaPrompt EngineeringReActTool-use
02
Freelance 2025

Customer Propensity
Pipeline.

Machine Learning · 607,056 records

Built a complete end-to-end ML pipeline on 607,056 records — ETL preprocessing, feature engineering (engagement, intent, and conversion scores), model training, 10-fold cross-validation, and deployment-ready CRM targeting recommendations. Applied K-Means (silhouette 0.843) and GMM (silhouette 0.891) segmentation; Logistic Regression and Random Forest classifiers achieved AUC > 0.93. Gradient Boosting regression hit RMSE 0.02084 — a 50% improvement over baseline. Outputs translated into tiered CRM strategies with full data governance documentation.

0.93AUC
0.891Silhouette
−50%RMSE vs. baseline
PythonXGBoostscikit-learnRK-MeansGMM
View Project →
03
Freelance 2025

Propensity Pipeline
at Scale.

Production MLOps · PySpark + Airflow + Docker

Re-engineered the propensity-to-purchase problem as a production-grade, distributed pipeline — moving from notebook to a containerised, orchestrated system. Built distributed ETL and feature engineering in PySpark to process the dataset at scale, with the full workflow orchestrated as an Apache Airflow DAG covering ingestion, transformation, training, scoring, and deployment. Packaged the entire stack in Docker for reproducible, environment-independent runs — demonstrating end-to-end MLOps and production deployment beyond the original analysis.

PySparkdistributed ETL
Airfloworchestrated DAG
Dockercontainerised
PySparkAirflowDockerMLOpsETLPython
View Project →
04
Wayne Trading Co. 2025

Global Sales
Story.

Tableau Dashboards · $12.6M Sales Data

Designed and executed a full data pipeline: SQL extraction → ETL transformation → EDA → Tableau visualisation across $12.6M in global sales (2019–2022), 7 regions, and 17 product categories. Surfaced APAC as the top-performing region ($3.58M) and identified tables as a loss-making category (−$62,888). Built 5 interactive Tableau story dashboards with drill-through capability; applied statistical analysis to prove higher discounts negatively correlate with sales volume — directly informing promotional strategy redesign.

$12.6Manalysed
7regions
5dashboards
TableauSQLEDAStatistical Testing
View Project →
05
Marriott International 2024

Loyalty
Research.

Statistical Research · Mixed Methods

Designed a full mixed-methods research framework for Marriott International — 5,000+ respondents at 95% CI with stratified random sampling. Applied Factor Analysis (PCA, Varimax), Cluster Analysis (Ward's Method), One-Way & Two-Way ANOVA, Multiple Linear Regression, and Chi-Square across Consumer, Corporate, and Home Office segments. Delivered data governance-compliant outputs including data quality checks and validated sampling methodology. Recommended a loyalty programme optimisation strategy validated across all three segments.

5K+respondents
95%confidence
3segments
SPSSFactor AnalysisANOVARegressionChi-Square
06

03 / Stack

A constellation
of tools.

Code

Python · SQL · R · pandas · NumPy · scikit-learn · matplotlib · seaborn · Jupyter

Models

XGBoost · LightGBM · Random Forest · Gradient Boosting · K-Means · GMM · Deep Learning · Neural Networks · TensorFlow · PyTorch

AI / NLP

Ollama · LLM Integration · Prompt Engineering · Agentic Workflows · Retrieval-Augmented Generation · LangChain-style Frameworks

Workflow

Feature Engineering · Cross-Validation · A/B Testing · Hyperparameter Tuning · SHAP · Model Interpretability · Model Monitoring · Data Drift

Deploy

AWS SageMaker · Streamlit · Dataiku · ETL Pipelines · DuckDB · Git · Docker · MLflow · API Integration

BI

Tableau · Power BI · EDA · SPSS · Excel · CRM · Data Governance · Quality & Integrity

04 / Journey

Eight years.
Four chapters.

2025 — 2026

Data Scientist

Freelance · Remote

Independently designed, built, and deployed end-to-end data science and AI solutions — managing the full project lifecycle from requirements and data engineering through to model deployment and client-ready deliverables. Applied ML, NLP, and agentic AI frameworks to solve real-world problems.

2023 — 2025

Senior Robotics & STEM Coach

Brainery Code · Singapore

Two years of daily practice in something data science values as much as Python: translating complex technical ideas to people who don't share your background. Taught 100+ students aged 6–16 across Python, Roblox Studio, LEGO Robotics.

2019 — 2022

Operations Manager

Stylar Digital Pte. Ltd. · Myanmar

Where I built my instinct for analytical thinking in a real business context. Used data analysis and trend forecasting to guide planning, budgeting, and resource allocation — translating raw operational data into recommendations leadership could act on.

2018 — 2019

Loyalty Programme Coordinator

Pan Pacific Hotels Group · Myanmar

The role that first got me thinking seriously about data. Owned the full loyalty programme lifecycle: CRM database architecture, ETL workflow design, governance. Contributed directly to 2,000+ pre-registrations.

05 / Foundations

Where it
was built.

2022 — 2025

BSc Data Science & Business Analytics

University of London · Singapore

Machine Learning · Statistical Modelling · Business Analytics · Database Management · Data Visualisation · Statistical Research Methods

2012 — 2016

BA Tourism Studies

National Management Degree College · Myanmar

Certified

Dataiku Core Designer · AWS MLA-C01 (in progress)

Dataiku Academy · AWS Certified Machine Learning Associate

Currently expanding — AWS Cloud Deployment · PyTorch · Deep Learning · Docker · Containerisation

Languages

EnglishProfessional
BurmeseNative
JapaneseConversational

06 / Contact

Let's build something.

Open to Data Science, ML Engineering, and Analytics roles — in Bangkok, Singapore, and globally.

bonoerikk@gmail.com
+65 9006 4760 github.com/Bo-Neau Bangkok / Singapore