Causal inference & uplift modeling
Estimating heterogeneous treatment effects and learning policies that optimize incremental outcomes rather than correlations alone.
Causal AI/ML Researcher · Hanoi, Vietnam
Studying cause, not just correlation.
I am Nguyen Duong Hieu, an undergraduate researcher at the National Economics University. My work sits at the intersection of causal inference and machine learning, with a focus on reliable decisions under distribution shift.
Research direction
Causal AI/ML.
I study how AI and machine learning can estimate causal effects, identify who benefits from an intervention, and support reliable decisions beyond correlation.
Causal representation & OOD
Learning representations that capture stable mechanisms and support robust prediction when environments or populations change.
Trustworthy & probabilistic AI
Building models whose assumptions, uncertainty, and failure modes can be inspected before they influence real decisions.
Experience
Research in context.
My experience spans causal ML, multimodal systems, academic networks, mathematical optimization, and teaching.
AI Intern, Fintech & Payment
VinSmart Future · Vingroup
- Developed an uplift-modeling proof of concept for targeted fintech marketing using Conditional Average Treatment Effects instead of conventional response probabilities.
- Built an experimentation pipeline on the Hillstrom and Criteo benchmarks, covering preprocessing, treatment-effect estimation, policy evaluation, and cost-effectiveness analysis.
- Designed a simulated A/B testing framework to evaluate whether learned targeting policies produce incremental outcomes under budget constraints.
Student Researcher
DAAI Lab · National Economics University
- Developed a multimodal workflow combining OCR and PhoWhisper ASR with Gemini for structured extraction from noisy visual and acoustic inputs.
- Built Sentence-BERT and BM25 hybrid retrieval with Reciprocal Rank Fusion over a 50,000-SKU catalog.
- Achieved F1 scores of 0.91 on visual inputs and 0.85 on acoustic inputs, with Recall@5 of 0.93.
Research Assistant
Does Network Position Cause Scholarly Success?
With Prof. Mike Nguyen (USC) and Prof. Minh Nguyen (FAU)
- Curated and structured large-scale academic collaboration networks to support causal analysis of how network position affects scholarly productivity.
Student Researcher
DATAOPT Lab · National Economics University ↗
- Implemented mathematical and convex optimization algorithms, reproduced experiments from the literature, and evaluated their behavior in machine-learning settings.
Teaching Assistant
Python for Data Science & Power BI · National Economics University
- Prepared practical exercises, explained programming and data-analysis concepts, and provided structured feedback on student submissions.
Publication
Published research.
My current publication applies behavioral modeling and structural equation analysis to sustainable consumption decisions.
2026
Promoting circular consumption behavior: Factors influencing Vietnamese Generation Z's intention to extend the lifespan of personal electronic devices
22nd International Conference on Socio-economic and Environmental Issues in Development · ICSEED 2026
Applied PLS-SEM to 733 respondents using an integrated Theory of Planned Behavior and Push-Pull-Mooring framework.
Selected work
Ideas made testable.
Beyond research prototypes, I build complete systems to understand how models behave when they meet real users and operational constraints.
FEATURED · 2026
DentFlow AI
An AI-assisted dental booking and clinic-flow platform that separates language intelligence from deterministic operational decisions.
- Built guarded AI intake and adaptive booking with structured outputs and deterministic fallbacks.
- Implemented capacity-aware slot selection and conflict-safe booking.
- Developed room scoring, service queues, route sequencing, realtime updates, and role-based access.
Next.jsTypeScriptSupabaseOpenAIVitestPlaywright
Soi Thuoc
A mobile-first reading assistant that helps Vietnamese seniors understand medication labels, bills, forms, and official documents from a single photo.
- Combined vision OCR, grounded follow-up Q&A, Vietnamese TTS/STT, and installable PWA support.
- Added validation, rate limiting, request IDs, privacy-conscious caching, and an API-key-free demo mode.
Next.jsVision OCRTTS/STTPWAAccessibility
Hanoi Temperature Forecasting
A five-day forecasting pipeline trained on more than ten years of weather data, with interpretable predictions and drift monitoring.
- Engineered 90+ lag, STL decomposition, and FFT features.
- Tuned a multi-output CatBoost model to R² ≈ 0.84 and deployed a Streamlit interface with SHAP and PSI/KS checks.
PythonCatBoostOptunaSHAPStreamlit
Academic profile
Training & toolkit.
A quantitative foundation in statistics and optimization, paired with the engineering tools needed to run reproducible experiments.
Education
BS in Data Science in Economics and Business
National Economics University
Relevant coursework: Machine Learning, Probability & Statistics, Optimization, Linear Algebra, and Data Structures & Algorithms.
Natural Science 1
Tran Phu High School for the Gifted
Technical skills
Research methods
Causal inference, CATE and uplift modeling, DAGs, experimental design, probabilistic ML, optimization, statistical analysis
Programming & data
Python, SQL, TypeScript, Pandas, NumPy, Scikit-learn
Machine learning
CatBoost, Optuna, SHAP, Sentence-BERT, BM25, LLMs, NLP, computer vision, OCR/ASR pipelines
Engineering
Git, Docker, GitHub Actions, Next.js, React, Supabase, Streamlit, Power BI, LaTeX
Recognition
Selected achievements.
Academic, research, and mathematical distinctions earned throughout my studies.
Third Prize · University-Level Scientific Research Competition, NEU
Top 10 · Hanoi Mathematical Modeling Competition
Academic Encouragement Scholarship (Excellent) · NEU
Outstanding Youth Union Member Award · NEU
Consolation Prize · Hai Phong City Mathematics Competition
IELTS 7.0
Contact
Let’s investigate a good question.
I am open to research assistantships, collaborations, and conversations around causal ML, reliable generalization, and applied AI systems.