HENG SENGTHAY
Y3 DATA SCIENCE // RUPP
CONNECT
RUPP FACULTY OF ENGINEERING DATA SCIENCE Y3 CAMBODIA

HENG
SENGTHAY

Third-year Data Science at the Royal University of Phnom Penh (RUPP). Architecting predictive time-series models, biomedical signal processing pipelines, and spatial remote sensing telemetry.

EXPLORE 6 PINNED PROJECTS → GITHUB @Thay32-Heng ↗ DOWNLOAD CV ↗ OPEN TERMINAL >_
28+ ML MODELS
14 DATA LAKES
99.2% TEST ACCURACY
Heng SengThay — head and shoulders portrait
HOVER TO REVEAL AURA

Currently

Data Science

Royal University of Phnom Penh (RUPP)

COHORT: Y3 (2026)
FACULTY: ENGINEERING // Dept. of Information Technology and Engineering

Featured Lab // Pinned 01

CODE ↗

ECG Signal Analysis

Biomedical DSP & QRS Peak Detection · SciPy, 2026

At a Glance // Telemetry

MODELS
28+
ACCURACY
99.2%
SERVING: FASTAPI + CUDA LATENCY: 2.4ms
Scroll to pull back
ENGINEERING PROTOCOL

END-TO-END DATA PIPELINE

From noisy multi-source ingestion to distributed feature extraction and sub-millisecond production inference.

01

INGESTION & DATA LAKE CURATION

Automated ingestion of multi-source telemetry, geospatial satellite datasets, and web streams into PostgreSQL & BigQuery data lakes.

APACHE KAFKA BIGQUERY
02

DISTRIBUTED FEATURE ENGINEERING

Matrix transformations, missing value imputation, polynomial expansions, and PCA dimensional reductions on cluster runtimes.

APACHE SPARK PANDAS
03

NEURAL TRAINING & HYPER-TUNING

Deep learning model training with Bayesian hyperparameter optimization, learning rate schedules, and cross-entropy loss minimization.

PYTORCH OPTUNA
04

CONTAINERIZED SERVING & MONITORING

Model packaging via ONNX runtime, asynchronous FastAPI endpoints, automated drift monitoring, and Dockerized edge deployment.

DOCKER FASTAPI
LIVE TELEMETRY

LOSS CONVERGENCE

250/250
EPOCHS
0.0142
TRAIN LOSS
99.2%
ACCURACY

ADAM OPTIMIZER CONVERGED WITHOUT OVERFITTING. REGULARIZATION: L2 WEIGHT DECAY (1e-4) WITH EARLY STOPPING.

ACADEMIC JOURNEY

FROM FOUNDATIONS TO INTELLIGENCE

From a January 2025 admission at RUPP to Cambodia’s flagship developer stages — the build log, in real repo dates pulled from GitHub.

2025
RUPP YEAR 1 // ADMISSION

DATA SCIENCE & ENGINEERING AT RUPP

20 JAN 2025 FACULTY OF ENGINEERING

Started the Data Science & Engineering programme at the Royal University of Phnom Penh — Python, statistical inference, data structures, linear algebra and discrete mathematics foundations.

2025
YEAR 1 // IEEE CAMBODIA

IEEE CAMBODIA TECHNICAL GATHERING

PARTICIPANT 10 MAR 2025 NICC // PHNOM PENH

Joined the IEEE Cambodia Subsection Landmark Community Gathering with students, developers and researchers, touring RUPP’s advanced laboratories alongside global engineering leadership.

  • Global tech standards — international engineering pathways, academic publishing and IEEE professional growth
  • Innovation & research — RUPP AI Innovation, Automation and Plant Tech laboratories
  • Ecosystem networking — university leaders, local builders and global technology pioneers
IEEE Cambodia Landmark Community Gathering group photo at NICC, Phnom Penh
2026
RUPP YEAR 2 // FIRST PUBLIC BUILDS

GITHUB: ANALYTICS SPRINT & FIRST APPS

JAN–MAR 2026 PYTHON // JAVASCRIPT // DART

Opened the public GitHub record in January 2026 and shipped the first wave: a February analytics sprint across sales, air-quality, gold-price and book-price data, ECG signal analysis, then JavaScript and Flutter mini-apps.

2026
COMMUNITY // VOLUNTEER

CAMBODIA BLOCKCHAIN SUMMIT

VOLUNTEER MAR 2026 – PRESENT PHNOM PENH

Serving as an event volunteer for the Cambodia Blockchain Summit community — a regional forum on digital public infrastructure, fintech innovation and decentralised architecture.

  • Event operations — on-site logistics and stakeholder tracking for hundreds of tech professionals
  • Delegate hospitality — welcoming international entrepreneurs, investors and digital-asset delegates
  • Cross-functional teamwork — rapid problem-solving with a fast-paced core team under tight deadlines
Speaking at a Cambodia Blockchain Summit community session with a microphone
2026
RUPP YEAR 2 // PRODUCT ENGINEERING

TYPESCRIPT PLATFORMS & COURSEWORK SYSTEMS

JUL–SEP 2026 SPARK // KAFKA // DOCKER // CI

Built and iterated the Ahnajak top-up platform line, Aureon AI and JamTlai in TypeScript, plus a Flutter churn-prediction app — alongside distributed data coursework in Apache Spark, Kafka event ingestion, Docker and automated CI/CD.

2026
RUPP YEAR 2 // APPLIED DATA SCIENCE

EDURISK, KASEKOR VISION & PRECIPITATION

SEP 2026 STREAMLIT // GEOSPATIAL // ARIMA

Published the applied data-science trio: EduRiskAnalysis student early-warning dashboard with 30 unit tests and GitHub Actions CI, the Kasekor-Vision Cambodian agri-data crop optimizer, and the 11-year Phnom Penh precipitation forecast evaluation.

2026
YEAR 3 // DEVELOPER CONFERENCE

CODE-C 2026 DEVELOPER CONFERENCE

PARTICIPANT 21–22 OCT 2026 CADT INNOVATION CENTER

Joined 3,000+ tech professionals at Cambodia’s largest developer conference to explore the tools, stacks and platforms shaping the AI era.

  • Next-gen tech stacks — agentic AI, open models and AI-powered software development
  • Industry networking — local and international developers, engineers and tech pioneers
  • Tech ecosystem — interactive brand exhibitions and masterclasses on future-ready infrastructure
CODE-C 2026 Developer Conference delegate group on the CADT Innovation Center stage
2026
YEAR 3 // NASA SPACE APPS

NASA SPACE APPS CHALLENGE PHNOM PENH

SHORTLISTED OCT 2026 – PRESENT CAMTECH // PHNOM PENH CHALLENGE 08

Shortlisted through the global registration screening and advanced with my team to the in-person Pre-Event Meetup in Phnom Penh — 100+ local innovators aligning teams and digging into NASA’s open space-data frameworks.

  • Advanced challenge selection — Challenge 08, AI-powered fire-safety insights from microgravity combustion data
  • Tech architecture & strategy — framework combining machine learning and computer vision over real NASA datasets
  • Agile team collaboration — role synchronisation across a cross-functional team before the main event
NASA Space Apps Challenge Phnom Penh 2026 delegate group photo
Competitors in the NASA Space Apps Phnom Penh challenge room
2026
YEAR 3 // GRADUATION CAPSTONE

MULTI-MODAL LLMS & EDGE INFERENCE

2026 – 2027 RAG // EDGE RUNTIME

Finalizing the RUPP Data Science Capstone: high-density vector database embeddings (RAG), sub-millisecond edge runtimes, and real-time predictive intelligence.

GITHUB PINNED REPOSITORIES // @THAY32-HENG

FEATURED RESEARCH LAB

Real-world data science systems, biomedical signal processing, agricultural remote sensing, and meteorological time-series predictive pipelines pinned on his official GitHub.

VISIT GITHUB PROFILE ↗
PINNED // 01 JUPYTER

ECG SIGNAL ANALYSIS

github.com/Thay32-Heng/ECG-Signal_Analysis ↗

Biomedical digital signal processing on electrocardiogram (ECG) waveforms. Implements baseline wander removal, bandpass filtering, Pan-Tompkins QRS complex peak detection, and cardiac arrhythmia pattern classification.

360 Hz
SAMPLING RATE
QRS PEAK
DETECTION
SCIPY
DSP STACK
PYTHON SCIPY SIGNAL PROCESSING
CODE ↗
PINNED // 02 JUPYTER

GOLD PRICE PREDICTION

github.com/Thay32-Heng/Gold-Price-Analysis---Simple-Prediction ↗

Financial econometric time-series regression and historical gold commodity price trend forecasting. Evaluates moving averages (SMA/EMA), exponential smoothing, rolling volatility metrics, and predictive regression errors.

XAU/USD
ASSET PAIR
MULTI-DAY
HORIZON
REGRESSION
MODEL TYPE
TIME-SERIES PANDAS SCIKIT-LEARN
CODE ↗
PINNED // 03 PYTHON

EDURISK ANALYSIS

github.com/Thay32-Heng/EduRiskAnalysis ↗

Student early-warning academic risk scoring dashboard built with Streamlit, Pandas, and Altair. Engineered transparent risk prediction rules, automated unit test suites with 30 passing tests, and GitHub Actions CI pipelines.

30 TESTS
UNIT TESTING
STREAMLIT
UI ENGINE
CI ACTIVE
WORKFLOW
STREAMLIT ALTAIR PYTEST CI/CD
DEMO ↗ CODE ↗
PINNED // 04 PYTHON

KASEKOR VISION

github.com/Thay32-Heng/Kasekor-Vision ↗

Empowering Cambodian agriculture with spatial data intelligence. Recommends optimal crop species, seed varieties, and planting density based on geographical coordinates, soil chemistry, and provincial climate factors via an interactive map.

CAMBODIA
TERRITORY
INTERACTIVE
MAP VIZ
AGRI-DATA
OPTIMIZER
GEOSPATIAL AGRI-TECH PYTHON FOLIUM
DEMO ↗ CODE ↗
PINNED // 05 JUPYTER

PHNOM PENH PRECIPITATION

github.com/Thay32-Heng/Phnom-Penh-Precipitation-Time-Series-Analysis ↗

11-year monthly precipitation time-series analysis and forecasting evaluation in Phnom Penh (2015–2025). Investigates seasonal monsoon variations, climate telemetry patterns, ARIMA/SARIMA lag structures, and predictive rainfall error bounds.

11 YEARS
2015 – 2025
PHNOM PENH
METEOROLOGY
TIME-SERIES
FORECASTING
METEOROLOGY PANDAS STATMODELS SEABORN
CODE ↗
PINNED // 06 JAVASCRIPT

PYFIRE WATCH

github.com/Thay32-Heng/pyfire-watch ↗

Engineered for the NASA Space Apps 2026 challenge: Harmonizing NASA MODIS and VIIRS satellite active-fire detections into a calibrated burning calendar, heat radiance clustering, and geospatial wildfire risk telemetry.

NASA APPS
SPACE HACKATHON
MODIS/VIIRS
SATELLITES
CALENDAR
CALIBRATION
NASA SPACE APPS REMOTE SENSING JAVASCRIPT
DEMO ↗ CODE ↗
REPOSITORY REGISTRY

ADDITIONAL PUBLIC REPOSITORIES

Explore additional utility libraries, mobile architectures, web engines, and data pipelines on his GitHub.

ALL 22 REPOSITORIES ↗
REPOSITORY NAME LANGUAGE / STACK DOMAIN / SCOPE ACCESS
JamTlai (ចាំតម្លៃ)
TypeScript / PWA Ultra-fast offline-first PWA price checker with dual retail/wholesale pricing in KHR for Cambodian stores VIEW ↗
Aureon-Ai
TypeScript / AI Autonomous agent runtime and interactive intelligent interface VIEW ↗
Churn-Prediction
Dart / Flutter Customer churn classification and mobile-enabled predictive telemetry VIEW ↗
AQ-Data-Analysis
Jupyter Notebook Air Quality index (AQI) data analysis, PM2.5 particulate modeling, and environmental time trends VIEW ↗
Books-Price
Jupyter Notebook Commercial pricing regression models and book market exploratory data analysis (EDA) VIEW ↗
Sales-Data-Analysis
Jupyter Notebook Retail revenue optimization, inventory cycle clustering, and seasonal sales forecasting VIEW ↗
TECHNICAL ARSENAL

SKILLS & CAPABILITIES

The stack I actually work in — built through coursework, lab reports and small data products that had to ship.

[ 01 ] DATA ANALYSIS

Cleaning, querying and visualising real datasets — from a messy CSV to a chart a decision can actually be made from.

  • Python
  • SQL
  • Pandas
  • NumPy
  • Matplotlib
  • Seaborn
  • Jupyter
  • Power BI
  • EDA

[ 02 ] MACHINE LEARNING

The maths underneath the models, and the models themselves: distributions, gradients, and a classifier that still holds up on unseen data.

  • Scikit-learn
  • TensorFlow / Keras
  • Probability & Statistics
  • Linear Algebra
  • Multivariate Calculus
  • Optimization

[ 03 ] SYSTEMS & BUILD

Getting the work out of the notebook — schema design, APIs, containers, version control, and the languages the apps are written in.

  • PostgreSQL
  • FastAPI
  • Docker
  • Git & GitHub
  • Streamlit
  • C++
  • JavaScript / TS
  • Dart / Flutter
COMMUNICATION UPLINK

GET IN TOUCH

Open for machine learning engineering collaborations, academic research, and high-performance data systems.

sengthay@rupp-data-lab: ~ (interactive shell) PORT: 443
sengthay@rupp:~$ system_init --identity
[OK] HENG SENGTHAY // YEAR 3 DATA SCIENCE (RUPP)
[OK] NEURAL CORE LOADED // 28 MODELS ONLINE
[OK] TYPE 'help' TO VIEW AVAILABLE COMMANDS.
sengthay@rupp:~$
DIRECT CHANNELS

ROYAL UNIVERSITY OF PHNOM PENH

Faculty of Engineering, Department of Information Technology and Engineering & Data Science and Engineering.
Russian Federation Blvd, Toul Kork, Phnom Penh, Cambodia.

PERSONAL EMAIL
sengthay32@gmail.com
RUPP SCHOOL EMAIL
sengthay.heng.3624@rupp.edu.kh
OFFICIAL GITHUB
github.com/Thay32-Heng
GITHUB @Thay32-Heng ↗ SEND EMAIL