@@ B.Eng in Artificial Intelligence @ FPT University | GPA: 8.62/10.0 @@
+ Published First-Author in Springer LNAI (AJCAI 2025, Q2)
+ 3 New Manuscripts Under Review (AJCAI 2026, ACM SIGKDD 2027)
+ Certificate of Merit — Honorable Student (Semesters 3, 4, 5 & 6)
+ Founder & Student Lead @ SpeedyLabX Research Group (10+ Researchers)
! Actively seeking AI Research Internships & Global Collaboration Opportunitiesidentity:
name: "Gia-Hung Nguyen Le (Hei)"
role: "AI Research Engineer & Undergraduate Researcher"
affiliation: "FPT University (Ho Chi Minh City, Vietnam)"
expected_graduation: "Dec 2027"
core_pillars:
- "Explainable & Trustworthy AI (SHAP, Integrated Gradients, Conformal Calibration)"
- "Time-Series Forecasting & Anomaly Detection (Transformers, GNNs)"
- "Multimodal Learning (Audio-Text Fusion, Graph Neural Networks)"
- "Efficient Speech Recognition (Low-Resource CTC ASR)"I am an AI researcher focused on developing robust, interpretable, and high-stakes machine learning models. As the founder and lead of SpeedyLabX, I mentor 10+ undergraduate researchers through rigorous experimental pipelines, paper reading workshops, and end-to-end conference submission workflows.
AJCAI 2025 • Published (Springer LNAI Q2) results = {
"MAE": 0.7470,
"R²": 0.9545,
"smoke_event_gain": "+57.7% vs baseline"
}Transformer-based PM2.5 forecasting framework integrating SHAP and Integrated Gradients for proactive health warnings.
|
AJCAI 2026 • Under Review benchmark = {
"dataset": "7 New Zealand Lakes",
"samples": "5,432 weekly causal benchmark",
"weighted_F1": 0.791
}Stress-testing multi-source water quality prediction under sparse monitoring with an 8-gate provenance-aware audit.
|
ACM SIGKDD 2027 • Under Review system = {
"framework": "Split-Conformal + Mondrian Pools",
"attribution_acc": "94.3% against ground truth",
"coreset_pruning": "10%"
}Distribution-free false alarm control and counterfactual attribution for look-ahead leakage prevention in time series.
|
Target: Elsevier Information Fusion (Q1) architecture = {
"audio": "Wav2Vec 2.0",
"text": "RoBERTa",
"fusion": "GATv2 Graph Attention"
}Dynamic conversational graph modeling temporal and inter-speaker dynamics with intrinsic attention explainability.
|
| Year | Venue / Target | Paper Title | Status & Role | Links |
|---|---|---|---|---|
| 2025 | AJCAI 2025 (Springer LNAI, Q2) | Proactive Air Quality Forecasting and Health Alert System for Melbourne | Published • First & Corresponding Author |  |
| 2026 | AJCAI 2026 | Stress-Testing Multi-Source Cyanobacterial Bloom Forecasting under Sparse Monitoring | Under Review • First & Corresponding Author |  |
| 2026 | AJCAI 2026 | Warning Before the Evening Peak: Health-Centred Air Quality Alerts in NZ & AU | Under Review • Second Author | |
| 2027 | ACM SIGKDD 2027 | Trustworthy Multimodal Anomaly Detection in Time Series with Conformal Calibration and Counterfactual Attribution | Under Review • Second Author | |
@inproceedings{le2025smokenet,
title = {Proactive Air Quality Forecasting and Health Alert System for Melbourne},
author = {Le, Gia-Hung Nguyen and Hoang, Gia-Bao Pham and Vo, Tan-Phat and Le, Thu and Nguyen, Nhu},
booktitle = {Australasian Joint Conference on Artificial Intelligence (AI 2025)},
series = {Lecture Notes in Artificial Intelligence (LNAI)},
publisher = {Springer Nature Singapore},
volume = {15496},
pages = {406--418},
year = {2025},
doi = {10.1007/978-981-95-4969-6_34},
note = {Q2 (SJR 2024)}
}
research_specializations:
time_series_and_tabular:
- Transformers (SmokeNet, PatchTST)
- Chronological CV & Leakage Auditing
- Multi-Source Sensor Fusion
explainable_ai_and_trustworthiness:
- SHAP & Integrated Gradients
- Conformal Prediction (Split-Conformal)
- Counterfactual Explanations
- Intrinsic GAT Attention Weights
multimodal_and_speech:
- Audio-Text Multimodal Fusion
- Graph Attention Networks (GATv2 / PyG)
- End-to-End CTC ASR (5.4M CNN-BiLSTM) |
technical_arsenal:
core_languages:
- Python (Expert)
- SQL / PostgreSQL
- LaTeX / Scientific Writing
frameworks_and_libraries:
- PyTorch & PyTorch Geometric (PyG)
- Hugging Face Transformers & Datasets
- Scikit-learn, OpenCV, TensorFlow
data_and_mlops:
- Weights & Biases (W&B)
- Docker & Containerization
- Linux, Git, CI/CD
- Pandas, NumPy, Matplotlib
- Streamlit (Rapid Prototyping) |
|
Languages
|
Deep Learning & GNNs
|
ML Ecosystem & XAI
|
MLOps & Tools
|
| Organization | Certification / Distinction | Focus Area |
|---|---|---|
| 🎓 FPT University | Certificate of Merit (Trimesters 3, 4, 5, 6) | Academic & Research Excellence |
| 🧠 DeepLearning.AI | Natural Language Processing Specialization | Attention, Transformers, QA Models |
| ⚙️ Coursera | Gradient to Production: MLOps & Model Serving | Model Serving, Drift, Deployment Pipelines |
| 💻 IBM | IBM Full Stack Software Developer | Microservices, Cloud Application Architecture |
| 🚀 Packt | AI Engineer Professional | Deep Learning Architectures & Production AI |
| 🐳 IBM | Docker, Kubernetes & OpenShift | Cloud-native containerized workflows |
const speedyLabX = {
name: "SpeedyLabX",
role: "Founder & Student Lead",
founded: 2025,
team_size: "10+ undergraduate researchers",
missions: [
"Bridge theoretical deep learning research with real-world explainability",
"Organize weekly technical workshops, code labs, and paper readings",
"Streamline publication workflows leading to international conference acceptances",
"Foster open-source benchmarks for environmental AI and speech processing"
],
domains: [
"Explainable AI (SHAP / Conformal Prediction)",
"Time-Series Forecasting & Anomaly Detection",
"Multimodal Learning (GNNs & Graph Attention)",
"Low-Resource Vietnamese Speech Processing"
]
};
